<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Daniel Burke on Medium]]></title>
        <description><![CDATA[Stories by Daniel Burke on Medium]]></description>
        <link>https://medium.com/@db_57288?source=rss-4a836b2efa74------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*GbfxayU1o3R2ey0b1cQXew.jpeg</url>
            <title>Stories by Daniel Burke on Medium</title>
            <link>https://medium.com/@db_57288?source=rss-4a836b2efa74------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 01:22:21 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@db_57288/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[AI for revenue growth: using ML  to drive more valuable pricing]]></title>
            <link>https://medium.com/data-science/ai-for-revenue-growth-using-ml-to-drive-more-valuable-pricing-89e8c790f795?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/89e8c790f795</guid>
            <category><![CDATA[management]]></category>
            <category><![CDATA[growth]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[revenue]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Tue, 04 Aug 2020 23:08:16 GMT</pubDate>
            <atom:updated>2020-08-06T13:27:16.351Z</atom:updated>
            <content:encoded><![CDATA[<h3>AI for revenue growth: using ML to drive more valuable pricing</h3><h4>How a national fitness brand used machine learning to optimize pricing and grow revenue by 11%</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*n5GOVxUwy2yS3PTTeIc9CA.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/@micheile?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Micheile Henderson</a> on <a href="https://unsplash.com/s/photos/growth?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><h3>Introduction</h3><p>Pricing optimization is a powerful lever for revenue growth, yet it’s too often put in the too-hard basket by too many companies.</p><p>This is because traditional pricing optimization methods can be both complex to implement and limited in their ability to accurately capture the full range of factors that can impact pricing.</p><p>Machine learning (ML) is well-suited to pricing optimization problems — both for its ability to handle complex features, as well as its ability to generalize to new situations. Moreover, recent advances in managed services has put these ML solutions within reach of virtually any organization.</p><p>In this anonymized example we explore how a company with no data science expertise was able to use managed ML services to implement an ML-powered pricing strategy that performed 2x above traditional approaches and resulted in estimated revenue growth of 11%.</p><h3>Situation</h3><p>FitCo is a premium fitness brand, based in Los Angeles, that operates a portfolio of over 600 gym and fitness center locations across the United States.</p><p>Having grown rapidly by acquisition over the past several years, management attention had now turned its attention to boosting organic revenue growth, which had been stubbornly flat on a per-studio basis.</p><p>FitCo had identified FitClass — its suite of specialty fitness classes — as a prime source of organic growth. Specifically, it had identified pricing of these classes as a major potential area of improvement.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dqsM6sPSHZ58UcDUJCvLCg.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/@shotsbywolf?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Geert Pieters</a> on <a href="https://unsplash.com/s/photos/gym-class-stretch?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>FitClasses are a popular offering across FitCo’s brands. They are premium experiences catering to niche fitness demand and sold on a pay-per-class basis on top of standard memberships.</p><p>Whilst FitCo had ensured a consistent user experience across its portfolio, local operators were still able to set schedules and prices for FitClasses in their studios with nearly total independence. As a result, prices varied widely between classes and locations.</p><p>Whilst FitCo understood that some of this variance reflected local conditions, they also suspected there was considerable room for improvement in the way prices were set across its portfolio.</p><h3>Complication</h3><p>FitCo had undertaken a pricing exercise two years ago in which prior management had opted to centralize FitClass pricing and institute a blanket price increase of between 10% and 20% across the board.</p><p>This blunt approach had not been successful. It had failed to take into account the price elasticity of demand of its customer set across the wide range of classes and locations, and the price increases actually resulted in overall revenue <em>declines</em> of 2% as the subsequent reduction in demand in many classes outstripped the increases in price. They were forced to unwind the price changes a couple months later.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/443/1*mBHml92Yn7g_tcymPU3b1g.png" /></figure><p>Though painful, that experience had at least been useful in giving FitCo a pretty solid dataset on price elasticity of its FitClass customer set. It could chart how class demand changed in response to price increases for each of three utilization bands — high (&gt;85%), medium (50–85%) and low (&lt;50%). Modeling out the various impacts of price increases on demand, FitCo estimated a revenue full potential of 15% from more efficient pricing.</p><p>To capture that higher revenue potential, FitCo now only needed to be able to accurately predict demand for classes into the future — whether current or new — to accurately model price impacts on revenue. This would enable FitCo to determine whether and how much of a price increase each class could profitably sustain.</p><p>FitCo had initially attempted this using traditional rules-based methods — effectively a series of if/then statements to set bands based on certain conditions. With some extensive trial and error they had managed to write a function that was estimated to generate about 5% in additional revenue. This wasn’t bad, but this approach had two primary limitations: (1) it failed to fully account for interrelationship of the wide variety of factors relevant to each class — it incorrectly predicted too many classes in the wrong band, causing a decline in usage — and (2) it failed to generalize to new class schedules or details at any given location — it couldn’t adequately account for new combinations of factors or scenarios.</p><p>Looking for alternative approaches, FitCo turned to ML.</p><h3>ML solution</h3><p>ML is well-suited to these types of classification problems precisely because of its ability to process a wide range of factors and generalize to unseen or new situations.</p><p>However, like most organizations of its size and in its industry, FitCo did not have ML capabilities or a team of data scientists on hand to design, build and deploy an ML solution. This had previously been a major barrier to ML adoption. Today though, the availability of ML managed services have largely democratized access to ML capabilities.</p><p>For their solution, FitCo chose Amazon SageMaker, which featured among other things, an AutoML capability called <a href="https://aws.amazon.com/sagemaker/autopilot/"><em>AutoPilot</em></a> that could take a simple tabular dataset and automate the process of building an ML workload around it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*74ZY05Gy-SXIbkae.png" /><figcaption>AutoPilot takes a simple tabular dataset and builds an ML workload around it</figcaption></figure><p>With <em>AutoPilot</em>, FitCo no longer needed a team of data scientists to get the benefit of ML. Instead, they were able to drive this initiative with a three-person project team consisting of the CFO as business owner, CTO as technology owner, and a single back-end developer responsible for building and integrating the solution.</p><h4>Training data</h4><p>To build their training dataset, FitCo gathered historical utilization data for each of their classes over the past two years.</p><p>Utilization for each class was expressed as a percentage of total places filled. FitCo converted the data in this column into ‘high’ , ‘medium’ and ‘low’ based on the utilization bands above and named this column ‘target’. This would be the column to be predicted by the ML model.</p><p>They then combined this data with a set of internal features they thought likely to indicate utilization. They also added a range of external data that they expected to be relevant. The result was a dataset with 800K instances and which contained the following features:</p><ul><li>type of class (categorical)</li><li>location (categorical)</li><li>day of week (categorical)</li><li>time of day (numerical)</li><li>instructor (categorical)</li><li>studio brand (categorical)</li><li>is a public holiday (binary)</li><li>is a school holiday (binary)</li><li>external temperature (numerical)</li><li>target (categorical)</li></ul><p>FitCo did some basic <a href="https://towardsdatascience.com/understanding-feature-engineering-part-1-continuous-numeric-data-da4e47099a7b">feature engineering</a> to better organize and format this data set, converted it to csv format and saved in an S3 bucket. They now had a dataset with which they could train their ML model.</p><h4>Amazon SageMaker AutoPilot</h4><p>FitCo chose Autopilot for its ability to simplify and streamline the core components of the machine learning process. AutoPilot automates the process of exploring data, engineering features, testing different algorithms, and selecting the best model. All it requires is that you provide a tabular dataset.</p><p>In addition, it automatically surfaces the code base it used, adding visibility and reproducibility into the process. This was an important differentiator for FitCo’s CTO because it gave FitCo the opportunity to explore, and learn from, the steps that had been taken to generate the model, as well as giving it a code base that it could modify and optimize it into the future.</p><p>To start the AutoPilot process, FitCo used the no-code interface available within Amazon SageMaker Studio. This required three key steps:</p><ol><li>Name their experiment</li><li>Point SageMaker to the s3 bucket where their training file was located, and</li><li>Define the variable to be predicted</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gsb2WqaCjFafBJmF7Z2t1Q.png" /><figcaption>AutoPilot’s no-code interface puts ML within reach of any organization</figcaption></figure><p>Once these details were entered, they simply hit Create Experiment and FitCo’s ML model build was underway, running a range of trials to determine the best performing ML approach.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QtPZZ0X4TlRL8jrXNKMjjw.png" /><figcaption>Complete list of trials run by AutoPilot to determine best performing model</figcaption></figure><p>This process took about an hour to complete. Once it had concluded, FitCo could then simply sort the list of trials to find the best performing model. They were able to generate the notebook that contained the code of this model, and they were also able to deploy the model to a SageMaker endpoint that enabled them to further test the inferences (predictions) generated by the model on new data, or even put it into production.</p><h3>Results</h3><p>When modeled against FitCo’s test dataset, the ML model outperformed their rules-based approach by 2x, increasing revenue by an estimated 11% overall.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/727/1*rINnAv0zw4yHlyfPaewowA.png" /><figcaption>The ML model outperformed rules-based approaches by 2x</figcaption></figure><p>This performance improvement stemmed primarily from the higher precision of the ML predictions vs the rules-based approach. In a multi-class classification problem, the challenge is not only to predict the right class, but to minimize the cost of inaccuracies. For example, it is less costly to inaccurately predict a low demand class as ‘medium’ than as ‘high’. Specifically with FitCo’s price elasticity profile, the cost of that particular error was approximately 4x greater in the form of lower demand.</p><p>A comparison of this performance between traditional and ML approaches can be seen below, and illustrates that the traditional approach actually outperformed ML in accurately predicting high demand classes. The issue is that it was unable to do so while <em>also </em>accurately predicting medium and low demand classes. Moreover, it made costly errors in inaccurately predicting low classes as high demand classes. The ML model was better able to more holistically map the shape of the data to account for both situations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/877/1*mjNhXz7hXYkFjzRUu4GVbA.png" /><figcaption>Rules-based approaches incorrectly predicted low classes as high in 25% of cases</figcaption></figure><p>These errors came at a considerable cost to revenue for the rules-based approach that the ML model could avoid. For example, inaccurately predicting a low demand class as high demand resulted in a demand decline of nearly 50%, far offsetting the 30% higher price paid by the remaining members.</p><p>The matrix below shows the revenue growth impact of each type of prediction, expressed as the difference in performance between traditional and ML methods.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/849/1*2RZFmvlgCzn6OlXO-If9Ew.png" /><figcaption>ML model was able to generate 5.5% higher revenue by more consistent accuracy overall</figcaption></figure><p>Though traditional methods managed to beat the ML model in correctly labeling a higher proportion of high classes as high (a metric known as ‘recall’), it did so by also inaccurately labeling many more ‘medium’ and ‘low’ demand classes as high too (known as ‘precision’).</p><p>As a result, although its accurate high predictions generated 1.4% higher revenue, that came at a cost of losing 2.8% of revenue from the demand declines of incorrectly charging higher prices to higher elasticity medium and low classes. A similar pattern emerged in low predictions; the ML model’s greater accuracy meant that it only reduced the price of classes for which one would expect to see higher demand.</p><p>As a result of this higher precision —the more accurate prediction of both high and low classes — the ML model was able to generate 5.5% higher revenue overall, more than double that of rules-based approaches.</p><h3>Conclusion</h3><p>Pricing optimization is a powerful lever for revenue growth, and the application of ML provides a powerful solution, which can often outperform traditional approaches.</p><p>In FitCo’s case, the application of ML to their challenge generated a 2x uplift in revenue growth vs their best performing rules-based alternative, and produced an estimated 11% uplift in revenue.</p><p>FitCo’s example helps demonstrate both how ML can be applied to optimize pricing, as well as the way managed services like SageMaker AutoPilot are able to put these powerful ML solutions within reach of virtually any organization.</p><p>Have an AI opportunity you’d like to explore? Get in touch with me on <a href="https://www.linkedin.com/in/dmhburke/">LinkedIn</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=89e8c790f795" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/ai-for-revenue-growth-using-ml-to-drive-more-valuable-pricing-89e8c790f795">AI for revenue growth: using ML  to drive more valuable pricing</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Stop leaving money on the table: using ML for revenue growth]]></title>
            <link>https://medium.datadriveninvestor.com/stop-leaving-money-on-the-table-using-ml-for-revenue-growth-78bfeba4180a?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/78bfeba4180a</guid>
            <category><![CDATA[business]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[chief-revenue-officer]]></category>
            <category><![CDATA[revenue-growth]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Tue, 26 May 2020 13:46:14 GMT</pubDate>
            <atom:updated>2021-09-08T08:30:20.204Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>AI for revenue growth: using machine learning to reduce customer churn</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rtFWpIsJ2n5wtg8mOJM26A.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/@micheile?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Micheile Henderson</a> on <a href="https://unsplash.com/images/things/money?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>According to a <a href="https://www.oreilly.com/radar/ai-adoption-in-the-enterprise-2020/">recent report</a>, fully 85% of companies are adopting some form of machine learning (ML) in their business.</p><p>However, in the same survey only 13% of these companies reported using ML in revenue-generating functions.</p><p>The fact that fewer than one in seven executives in revenue-generating roles such as Chief Revenue Officer (CRO) or Head of Sales are using ML suggests that too many of these executives regard it either as not relevant, or technologically out of their reach.</p><p>Yet, neither of these things is true.</p><p>In this anonymized example, we’ll see how a mid-market industrial services company, with no ML expertise, was able to use ML to restart revenue growth and ultimately add 30% to its enterprise value.</p><p>This example helps illustrate the ways in which ML is an important technology for revenue generating executives to be using to accelerate revenue growth and increase enterprise value.</p><h4>Company context: <em>ServCo</em></h4><p><em>ServCo</em> is a $50M revenue industrial services company, based in Ohio, which provides industrial equipment maintenance and repair services to customers across the US Midwest.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YrqVlqTyoyKGP0mpprbing.jpeg" /><figcaption>ServCo provides industrial equipment services</figcaption></figure><p>After an acquisition by private equity, <em>ServCo</em> undertook a business rationalization process to improve efficiencies and shift to higher margin value-added services.</p><p>Whilst this was successful in improving EBITDA, top line growth remained stubbornly flat.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_1uUj47J5mun84uD47zwQg.png" /><figcaption><em>ServCo revenue remained flat</em></figcaption></figure><p>Analysis conducted by <em>ServCo’s</em> CRO identified contract renewals, or churn, as the primary inhibitor of growth, with about 15% of <em>ServCo’s </em>1200 customer accounts failing to renew in a given year.</p><p>At this rate, customer churn was costing <em>ServCo </em>about $7.2M in fewer customers and lost revenue, in the next year alone.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ae_6Os_6RhmcDOiyQVZROw.png" /><figcaption>Customer churn costing $7.2M in lower revenue in next year alone</figcaption></figure><p><em>ServCo</em> had traditionally relied on broad-brush tactics to reduce churn, such as ad-hoc discounts, and other engagement methods like periodic webinars and automated emails.</p><p>However, with pressure mounting to improve performance, ServCo turned to ML to help kick-start revenue growth.</p><h4>Using ML to predict churn</h4><p>Predicting churn is a classification problem for which ML is well-suited. It works by analyzing historical data to learn the patterns that are most predictive of churn.</p><p><em>ServCo </em>soon learned that they didn’t need much to get started. They put together a small ‘tiger team’, of one junior developer, Freya, who took hands-on responsibility, overseen by the CTO and the CRO as the technical and business owners respectively.</p><p>Freya’s first task was to find suitable customer data that described past instances of churn. Freya was able to compile about three years of customer data in a dataset that contained ten characteristics (‘features’) of those customers, including whether or not they had churned (the target variable). This data set comprised a total of ~3K rows (‘instances’). Not a huge dataset, but enough to get started.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1lv1yMIIhi5R3YWo_qdwyg.png" /></figure><p>Her next task was to use this data to train a model that could help predict churn.</p><p>Like most companies in their industry, <em>ServCo</em> had no in-house ML expertise, so they turned to <a href="https://aws.amazon.com/sagemaker/"><em>Amazon SageMaker</em></a>: a managed ML platform that makes it easy for any developer to build, train and deploy their own ML models, with no data science expertise required.</p><p>With a little research (aka she googled it), Freya identified XGBoost as a good all-purpose algorithm to use for this classification task, which was also one of the 18 algorithms supported natively within SageMaker. <em>ServCo</em> uploaded their data into an S3 bucket, opened a SageMaker notebook, and got to work by defining their algorithm in SageMaker.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GUu9GmiRH4e2aKUmJ0wVyA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1Vg0uAlzgQqx8i6PvG8iRg.png" /><figcaption>XGBoost is natively supported in SageMaker</figcaption></figure><p>The process of turning an algorithm into a model is called training, and is the most important — and most daunting — part of an ML workload. This is the process by which an algorithm runs over your training data to identify patterns that explain the data and ultimately make predictions.</p><p>To run this training process, <em>ServCo</em> made use of SageMaker’s built in hyperparameter optimization (HPO) function, which automated the selection of hyperparameters to the find the best combination based on the specifics of their dataset.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Jv4eWIdvxUWPiUsUTkMz1A.png" /><figcaption>HPO automates the most difficult parts of ML</figcaption></figure><p>With this defined, all that was left was to call fit on their ‘tuner’ to kick-off the training process, and SageMaker started the process of comparing 20 different variants to find the best performing version of their model. Using a C5 EC2 compute instance for training, this process only took 20 mins.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ekJyjv3e4eWhBMhJw9-gRg.png" /><figcaption>Calling .fit() kicks off the training process</figcaption></figure><p>Once finished, it was time to see how it performed. To do this, <em>ServCo</em> used SageMaker’s built-in deployment function to make predictions against its best performing model. This enabled it to test the model’s performance on the 20% of its training dataset they had reserved for this purpose.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mXjx6a7pLSkGPmroE0Pb3w.png" /><figcaption>Calling .deploy() creates an endpoint for testing</figcaption></figure><p>And the results looked pretty good! The model was able to accurately identify churning vs non-churning accounts 94% of the time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6rmiZxp0GnsHIIcllqxM9A.png" /><figcaption>ServCo’s first ML model captured 65% of churn</figcaption></figure><p>To see how useful the model would be, <em>ServCo </em>also looked at the model’s ‘confusion matrix’ — essentially a comparison of false positives and false negatives. On these deeper metrics, known as recall and precision, the model also performed well: its predictions accurately captured 65% of all churning accounts, and did so with 94% precision.</p><p>Whilst there was much more refinement that could further improve those metrics, this was enough to get started. And with that, in just a couple hours of development time, and with no specialized expertise, <em>ServCo</em> had built something that previously took highly specialized teams weeks to achieve: they had trained and deployed an ML model that could predictively identify churn for a large majority of their customer set.</p><h4>Applying ML to grow revenue</h4><p>To understand the value of this ML model, it’s important to first understand the challenges of traditional churn mitigation strategies.</p><p>To do this, we can consider <em>ServCo’s</em> most recent churn mitigation campaign, in which they’d targeted accounts to receive churn mitigation attention according to broadly observed characteristics such as reduced spend, increased customer service contacts, lower customer satisfaction, or simply the fact that they have been a customer for a long time.</p><p>This segmentation identified the following three customer segments as likely to churn: 1/ those who’ve had recent service issues, measured by the top 15% in customer service contact growth; 2/ those who appeared generally dissatisfied, measured by the lowest 15% customer satisfaction scores; and 3/ those who’ve reduced the number of services they’re using in the past year, measured by the top 15% reduction in service use.</p><p>Accounting for overlap, this amounted to ~30% of their customer base, totaling 362 of their 1200 accounts. To keep these efforts economical, the total budget was set at 1% of revenue. This meant they could spend just over $2K per targeted account, or 5% of their average customer’s spend.</p><p>Now, let’s say that <em>ServCo’s</em> segmentation was pretty accurate, and that they were able to identify churning accounts at 50% higher accuracy than a random sample.</p><p>Further, let’s say that their spend (say in credits, discounts or tailored attention) was pretty effective at reducing churn — and that this greater attention (equivalent to 5% of average customer spend) was able to successfully prevent 20% of churning accounts.</p><p>On this basis, their 362 targeted accounts would capture 45% of actually churning accounts, totaling 81, from which they would be able to convert 20%, resulting in a total of 16 ‘saved’ accounts — a reduction in churn of 1.4 percentage points, from 15% to 13.6%.</p><p>This creates the ‘standard’ churn mitigation campaign shape below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iPmdZ52DzmpVmu3qvdAmsg.png" /></figure><p>Now let’s compare this to an approach using ML.</p><p>To apply ML, <em>ServCo </em>simply feeds its customer data into its churn model, and uses its predictions to identify the individual accounts that should receive churn mitigation attention.</p><p>Based on the performance metrics of <em>ServCo’s</em> model when trained on historical data, we would expect it to correctly capture 65% of churning accounts with 94% accuracy rate.</p><p>Applied to ServCo’s full customer base, this means the model will identify 124 accounts as likely to churn. These become the accounts that would receive the churn mitigation attention outlined above.</p><p>Whilst 124 accounts is far fewer than the 362 accounts <em>ServCo </em>previously identified, the accuracy of the model is far higher accuracy (94%). This means that we can expect 117 of those accounts to be correctly identified as actually churning accounts — a full 25% <em>more</em> than identified by traditional segmentation, with only one-third the effort.</p><p>This efficiency pays off in terms of effectiveness; it means that the spend per account can be far higher without costing more overall. In fact, because of this efficiency, even while keeping the overall budget the same, <em>ServCo </em>can now afford to spend three times more per action on these targeted accounts than the traditional approach, enabling it to deliver far more valuable, high-touch or tailored assistance to customers predicted to churn.</p><p>With this higher effectiveness, not only is <em>ServCo </em>now able to identify more churning accounts, but they’re successfully influencing a higher proportion as well.</p><p>Thus the shape of churn reduction efforts using ML looks entirely different, with both greater efficiency and effectiveness on its side, ServCo is correctly identifying more accounts (117) and converting more (82), resulting in a churn reduction of about 7ppt, ~5x higher than with traditional methods.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SHv9xnmTnPEIU76qO6SoMg.png" /><figcaption>ML reduces churn 7ppt; 5x standard approach</figcaption></figure><p>A reduction of 7ppt, halves <em>ServCo’s </em>churn rate, from 15% to 8%.</p><p>This reduction in churn has a substantial impact on revenue growth, adding $3.8M in additional revenue from the current year, which is fully $2.7M more than would have been seen under the ‘standard’ churn campaign. Altogether this transforms 0.3% historical growth into 8% growth.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tbJSsiuxZBKxaMf6HgORJA.png" /><figcaption>ServCo revenue growth jumps to 8%</figcaption></figure><p>Moreover, as this growth was achieved within the same budget as already reserved for traditional methods, this revenue growth translates to meaningful EBITDA improvement, generating $13.9M EBITDA vs $10.6M baseline expectations and $11.2M standard churn expectations.</p><p>These increases in revenue and EBITDA have a considerable impact on <em>ServCo’s </em>potential Enterprise Value. However, it’s also worth noting the valuation benefits of ML adoption itself, quantified in recent Harvard Business Review <a href="https://hbr.org/2019/06/research-investors-reward-companies-that-talk-up-their-digital-initiatives">research</a> which found that companies who adopted advanced technology could expect to receive ~3%–9% higher valuations on average than those that had not. This higher valuation is reflected in <em>ServCo’s </em>higher multiple below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bzGSa0NGnbreS9TllJ8BZQ.png" /><figcaption><em>ServCo’s enterprise value has jumped by 30%</em></figcaption></figure><p>Altogether, the additional EBITDA and higher multiples generated by the adoption of ML increases <em>ServCo’s </em>potential Enterprise Value to $147M, an increase of 30% compared to standard churn mitigation approaches (from $113M to $147M).</p><p>Furthermore, its ability to more accurately and effectively identify and mitigate churn has enabled it to jump-start revenue growth from its historical levels of 0.3% to 8%.</p><h4>Conclusions</h4><p>This is a simple but useful example of<em> </em>how CROs and other revenue-generating executives can apply ML to deliver meaningful value.</p><p><em>ServCo’s </em>adoption of ML helped jump-start revenue growth and added 30% to its enterprise value. In doing so they helped demonstrate just one way in which ML solutions can be a powerful lever for growth. Moreover, their story illustrates how ML is well within reach of virtually any company, irrespective of technological sophistication.</p><p>This is an important lesson for CROs who still regard ML as not relevant to them, or as too far out of reach.</p><p>As ML continues to be rapidly adopted by companies like <em>ServCo</em> all around the world, CROs and other executives who continue to ignore its potential, risk increasingly being seen as leaving money on the table.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=78bfeba4180a" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/stop-leaving-money-on-the-table-using-ml-for-revenue-growth-78bfeba4180a">Stop leaving money on the table: using ML for revenue growth</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[(Successfully) Crossing the Chasm with AI]]></title>
            <link>https://medium.datadriveninvestor.com/successfully-crossing-the-chasm-with-ai-ae88a0d19eb6?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/ae88a0d19eb6</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[enterprise-technology]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Tue, 29 Jan 2019 13:42:15 GMT</pubDate>
            <atom:updated>2019-02-12T17:35:32.072Z</atom:updated>
            <content:encoded><![CDATA[<figure><a href="http://www.track.datadriveninvestor.com/1B9E"><img alt="" src="https://cdn-images-1.medium.com/max/700/0*BqeTGWCAmy8OCB_6" /></a></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/0*qjbOwutzN4iISEcP.gif" /><figcaption>Codifying the Do’s and *D’OH!* n’ts of AI product development</figcaption></figure><p>The year 1991 is remembered for many things: Operation Desert Storm, Michael Jackson’s <em>Dangerous</em>, the box office success of <em>Terminator 2</em>.</p><p>It is not remembered for major advances in AI.</p><p>Indeed, 1991 occurred during the darkest depths of the <a href="https://en.wikipedia.org/wiki/AI_winter">AI winter</a>, a period in which AI technology struggled through a multi-decade period of disinterest from investors, researchers and the general public alike.</p><p>Yet in 1991 a book was released which has re-emerged as an astonishingly prescient playbook for AI product development in 2019.</p><p>That book is <a href="https://www.amazon.com/Crossing-Chasm-Marketing-High-Tech-Mainstream/dp/0060517123"><em>Crossing the Chasm</em></a>, and the lessons contained within it provide three clear principles worth understanding by anyone looking to drive enterprise adoption of AI technology today.</p><h3>The Cliff notes on Crossing the Chasm</h3><p>The central tenet of <em>Crossing the Chasm</em> is that the marketing of disruptive technology needs to be tailored to each stage of the technology adoption life cycle.</p><p>The critical insight of the book is that there is one jump in particular that every disruptive tech product must make in order to survive. That is the jump between Early Adopters and the Early Majority.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k0MUJU5EOwprrRNfhIOtDA.png" /><figcaption>The Chasm is the gap between the Early Adopters and the Early Majority (source: <a href="https://www.google.com/search?q=crossing+the+chasm+images&amp;tbm=isch&amp;source=iu&amp;ictx=1&amp;fir=bMbOFbVIlvmAYM%253A%252CP8dutS-RhC77zM%252C_&amp;usg=AI4_-kRe_SFifsIp6rD64OE_1Bv51hZfXg&amp;sa=X&amp;ved=2ahUKEwi4wPTh-5HgAhWGZd8KHZEwA98Q9QEwAHoECAAQBA#imgrc=bMbOFbVIlvmAYM:">Prototypr</a>)</figcaption></figure><p>This is the eponymous ‘chasm’. It is the gap between visionary early adopters and the pragmatic early majority. And this is what makes this book so relevant to AI in 2019 because it is exactly where we are with regards to the adoption of AI technology.</p><p>There are two particularly salient points here worth pausing on before we dive into the specifics of AI adoption.</p><p>The first is that customers, rather than technology, drive the technology adoption lifecycle.</p><p>The determining factor for each group — from Innovators, to Early Adopters, Early Majority, Late Majority and ultimately Laggards — has little to do with the technology itself and almost everything to do with the reasons that customers pursue (or more commonly, resist) adoption. In other words, it is a profoundly customer-centric framework.</p><p>The second point is that the reason that the ‘chasm’ occurs is that this is the first time in the product’s life that, for its next customer set, what has been its relative strength — its new technology — becomes its relative weakness.</p><p>This is critical to understand. Earlier cohorts like Innovators will tend to adopt new products purely <em>because</em> of new technology in it; the novelty itself has value. Whereas the pragmatic Early Majority — while not necessarily opposed to technological novelty — are not driven by it. What drives them is usefulness. And when you care solely about usefulness, ‘new’ can often simply mean ‘unproven’.</p><h3>What are the lessons for AI product development?</h3><p>This is exactly where we are with AI today. We’re past the Innovator phase, which means that AI products must now be designed and marketed with this Early Majority in mind.</p><p>Applying the lessons of <em>Crossing the Chasm</em>, three very clear principles emerge for AI product proponents to prioritize in order to get their AI products successfully to the other side:</p><h4>1. Demonstrate ROI that the Early Majority can feel 🤩</h4><p>The most successful AI products of this era will those establish an undeniable link between the adoption and an increase in ‘Meaningful ROI’.</p><p>Meaningful ROI is a term we use to describe a business impact that is both measurable and tactile. It is commonly referred to as the source of ‘pain’. Indeed, Geoffrey Moore calls the most receptive Early Majority customers: ‘<a href="https://www.forbes.com/sites/danschawbel/2013/12/17/geoffrey-moore-why-crossing-the-chasm-is-still-relevant/#1865a08e782d">pragmatists in pain</a>’.</p><p>The best early example of this dynamic is the early adoption of AI products within enterprise sales and marketing functions. The pain these teams feel comes from the increasingly crowded sales environment in which they operate. The proliferation of new marketing channels combined with the fall in the cost of entry has produced a customer set whose attention is saturated. It is increasingly hard to cut through, meaning more wasted effort, and more missed sales targets.</p><p>The solution to this problem, according to <a href="https://blog.topohq.com/account-based-marketing-the-definitive-guide/">over 90% of B2B marketers</a>, is better personalization; being able to take tailored actions in relation to the right customer at the right time, commonly known as Account-Based Marketing (ABM). ABM is very effective, purported to deliver up to <a href="https://flipmyfunnel.com/flipping-funnels-weekly-account-based-marketing-isnt-the-death-of-anything/">3x revenue uplift</a> vs standard methods. The problem is that traditional ABM doesn’t easily scale. By necessity, its high touch requirements have always confined it only to a handful of the most important accounts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/900/1*GrbTer7xL55RuS-upPJ9lQ.png" /></figure><p>AI is tailor-made to solve this problem, which is why we’ve seen the very effective entry of AI products like Lattice Engines, a market leader in the fast-growing predictive revenue generation sector, using predictive analytics to transform revenue generating capabilities within organizations, including finally enabling ABM at scale.</p><p>Lattice’s AI is the application of proprietary machine learning technology to a combination of first- and third-party customer data to predictively identify patterns within an existing and potential customer base, identifying which customers are most likely to respond to specific actions at specific times. The application of AI to this data set has not only reduced the pain of scaling ABM, but is delivering measurable ROI in terms of <a href="https://intelligentgrowth.siriusdecisions.com/studies/2017-state-of-abm-study-findings">larger deal sizes</a>, <a href="https://hbr.org/2015/03/making-the-consensus-sale">higher conversion rates</a>, and <a href="https://www.salesfusion.com/resource/10-key-account-based-marketing-stats-marketers-know/">faster revenue growth</a>.</p><p><strong><em>The lesson of Lattice for those looking to Cross the Chasm?</em></strong> Find the undeniable connection between the adoption of your product and Meaningful ROI, and lead with that message.</p><h4>2. Drive predictions through to actions 🔥</h4><p>Successful Early Majority AI products will be those that drive to action.</p><p>The tendency for AI products is to lean on the power of prediction: ‘you can know things you never used to know’. This is no longer enough.</p><p>The pragmatism of Early Majority customers mean they will require AI products to make it easier for them to ‘do’ something, not just easier to ‘know’ something.</p><p>This has driven the success of vertical AI applications; applications focused on specific industries.</p><p>Recruiting is a promising example of this. The previous iteration of recruiting technology focused primarily on expanding the total applicant pool for any given job posting, a process which it has done so successfully that over 50% of talent acquisition leaders nominate <a href="https://ideal.com/ai-recruiting/">sorting through the deluge of applications</a> they receive as the hardest part of the recruitment process.</p><p>Enter AI, through companies like Ideal AI, which is applying techniques like unstructured data processing and reinforcement learning algorithms to automate the recruitment screening process in a way that not only reduces but often removes the need for human intervention in consuming activities like manually screening resumes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QqDtlDWQ-P1CQxrPB3hDrw.png" /></figure><p>The key to Ideal’s success lies in the application of its AI all the way through to the actual human benefit.</p><p>Ideal’s core technology gives it the ability to process unstructured resume data and use machine learning to predict the suitability of a candidate to a particular role. However, its <em>value</em> to Early Majority consumers lies not in that technology, but in its ability to translate that technology to action, by integrating into existing applicant tracking software to automatically shortlist the best candidates and enable easy outreach, as well as other value-adds like automatically identifying other candidates from different pools that have previously applied, and suggesting outreach to them as potential fits.</p><p><strong><em>The lesson of Ideal AI for those looking to Cross the Chasm? </em></strong>Push the application of your technology as far as it can go towards reducing or removing human action and highlight that functionality as much as possible.</p><h4>3. Provide clear user controls 💪</h4><p>The final principle for AI product proponents to prioritize is that the most successful AI products will be those that keep the user firmly in control.</p><p>This can be summarized as the ‘anti-black box’ principle and stems from the fact that Early Majority customers will be generally more skeptical of new technology, particularly if they feel like they are ceding any kind of meaningful control.</p><p>The field of ‘AI-UI’ is an exciting growth area of user experience and design, but this principle extends beyond design features like better dashboards. Instead it goes to the ability for users to review and adjust the actual performance and outcomes of the AI models in use.</p><p>A great example of this comes in the field of AI-enhanced legal services, in which companies like <a href="https://www.chisel.ai/government.php"><em>Chisel AI</em></a> operate to automate important but time-consuming and repetitive tasks like removing confidential information from (redacting) legal documents.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/810/1*exwKltazWZAEEbSk7Aoq1g.png" /></figure><p>Chisel uses AI to process a vast variety of text documents, and automatically identifies the elements of a document that are likely to be confidential — automatically presenting a redacted version of that document to a human user.</p><p>Chisel has added two critical elements to its workflow design to promote its adoption among Early Majority users: the first is that Chisel provides an easy opportunity for humans to review every redaction (and ignored redaction) by providing a simple summary containing each redacted portion of the document; and the second is that users are able to adjust the sensitivity of the redaction algorithm at any time from within the application, dialing up or down the sensitivity of the prediction algorithm, and thereby affecting the likelihood that the algorithm will flag potentially confidential information within the document.</p><p>Each of these measures — having built-in human-in-the-loop review mechanisms, as well as enabling easy user control over the application of the algorithm — reduce the raw effectiveness of the technology by requiring more human input (<a href="https://www.canadianbusiness.com/innovation/knote-document-redaction-ai/">Chisel still claims cost reduction of 80%</a>), but taken together they drastically streamline the likelihood of adoption of by Early Majority users because they effectively ameliorate potential concerns around loss of control.</p><p><strong><em>The lesson of Chisel AI for those looking to Cross the Chasm?</em></strong> Prioritize product design features that keep the user in control, even if they reduce the raw impact of the technology.</p><h3>Conclusion</h3><p>The first batch of enterprise AI products are fast approaching (or are already in-flight over) The Chasm. Few have yet made it successfully onto the other side.</p><p>Learning the 30-year-old lessons of Crossing the Chasm provides an unlikely but very clear set of principles that AI product proponents should prioritize as they seek to push forward the adoption of AI products either within their organizations or across the broader market.</p><p>Applying these principles will give AI products the best chance of successfully making the jump from enthusiastic Innovators to the prosperous Early Majority, and will deliver what very well may be the most consequential phase of AI product adoption we’ve seen to date.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HSy3Wk9FAKxvkVxMqdiUyA.png" /></figure><p>At Asta, we’re applying predictive AI to transform the flow of information at work. You can download it right <a href="https://chrome.google.com/webstore/detail/asta/gleedeafjlflbnjkopdccnnjaegicfgh/">here</a>.</p><p>And you can also follow me directly on <a href="https://www.linkedin.com/in/dmhburke/">LinkedIn</a> and <a href="https://twitter.com/danielmhburke">Twitter</a> for more about our AI future.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NxcH9o5IuT4RGMgJc8NuaQ.jpeg" /></figure><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fupscri.be%2Fb2a0d6%3Fas_embed%3Dtrue&amp;dntp=1&amp;display_name=Upscribe&amp;url=https%3A%2F%2Fupscri.be%2Fb2a0d6%2F&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=upscri" width="800" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/0707f5c806284d01a4a13c7b13a91ce3/href">https://medium.com/media/0707f5c806284d01a4a13c7b13a91ce3/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ae88a0d19eb6" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/successfully-crossing-the-chasm-with-ai-ae88a0d19eb6">(Successfully) Crossing the Chasm with AI</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[This technology will make 2019 the most impactful year for work in a generation]]></title>
            <link>https://medium.datadriveninvestor.com/this-technology-will-make-2019-the-most-impactful-year-for-work-in-a-generation-c092414ac8ee?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/c092414ac8ee</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[enterprise-technology]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Fri, 11 Jan 2019 14:44:46 GMT</pubDate>
            <atom:updated>2019-01-25T19:48:31.944Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*yCEMMcdNtRuBGGaQgeKnJQ.jpeg" /></figure><p>The past couple decades have been pretty bleak for enterprise technology.</p><p>While work has gotten faster, more flexible, and more frenetic, the technologies that power our work have remained resolutely the same.</p><p>Fortunately, 2019 looks like the year this finally changes.</p><p>An exciting new wave of enterprise technologies is about to redefine the way we work, and they’re all powered by the same transformative technology.</p><p>That technology is <strong>predictive analytics</strong>.</p><p>Predictive analytics already underpins the vast majority of the tech applications that dominate our lives: Google, Amazon, Facebook, Uber, Netflix .</p><p>Three distinct trends have all come together to make it likely that 2019 is the year that this technology finally transforms the way we work. And if that is the case, 2019 will likely be the most impactful year in enterprise technology that we’ve seen in a generation.</p><h3><strong>But first, what is predictive analytics…</strong></h3><p>Predictive analytics simply refers to the use of <a href="https://www.techrepublic.com/article/predictive-analytics-the-smart-persons-guide/">historical data to predict future events</a>. In this sense, it’s not new. Humans have been doing this in one form or another pretty much forever.</p><p>What’s changed — and what continues to change more rapidly each year — is the ability for technology to improve both the scope and accuracy of those predictions.</p><p>This improvements we’ve seen in this field even in the past five years has been genuinely life-changing.</p><p>Much of your life is already governed by predictive technology. If you get up in the morning and ask <a href="https://www.forbes.com/sites/bernardmarr/2018/10/05/how-does-amazons-alexa-really-work/#4d9a8bc61937">Alexa for the weather</a>, open <a href="http://ltd.edc.org/big-data-driving-google-maps">Google Maps to check the traffic</a> on your route to work, <a href="https://www.predictiveanalyticsworld.com/patimes/forecasting-uber-recurrent-neural-networks/8755/">order an Uber</a> to get there (and baulk at the surge pricing), or absent-mindedly <a href="https://www.predictiveanalyticsworld.com/patimes/predictive-analytics-facebook-a-love-story0823151/6117/">check your Facebook feed</a> while you wait for it to arrive, you’ve already been the beneficiary of predictive technology.</p><p>Predictive technologies have transformed nearly every major part of your life over the past decade, with one notable exception: your work.</p><p>While consumer technology has jumped ahead in leaps and bounds over the past two decades, enterprise technology — the technology that supports the way we work — has been left behind. This is a historical anomaly. Since the dawn of technology itself, and certainly since the industrial revolution, work was reliably the place you expected to see the first waves of technological innovation play out.</p><p>Yet despite the drastic changes in the way we work over the past couple of decades, very little about the technology supporting your work has incorporated predictive technology: you still do your own scheduling, write your own emails, search for your own information, respond to mundane and repetitive requests.</p><p>There are three trends have come together that set to make 2019 the year that work finally gets predictive.</p><h3><strong><em>1. The information crisis at work has reached a tipping point</em></strong></h3><p>It’s a <a href="https://www.executiveforums.com/single-post/2017/03/03/Information-Overload-What-is-it-doing-to-your-employees">well-worn point</a> that the internet age throws off more information than we reasonably know what to do with.</p><p>At the turn of last century, the amount of information available to the average person doubled roughly every 500 years. By the turn of this century that had jumped to doubling every two years. Today we double the amount of information we have access to roughly every month.</p><p>Unsurprisingly, nowhere is this more acute than at work. As the economy continues to transition to a <a href="https://en.wikipedia.org/wiki/Knowledge_economy">knowledge economy</a>, more and more jobs rely on the ability to process and act on information. However, the sheer volume of actionable information that the average worker has to deal with is getting out of hand. According to a <a href="https://www.mckinsey.com/industries/high-tech/our-insights/the-social-economy">McKinsey report</a>, today’s workers now spend around 20% of their time just processing the various streams of information directed at them, just to do their jobs.</p><p>In effect, today’s businesses are paying employees for five days and only getting them for four.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yryzuml28gfl7k2WVOBdBQ.jpeg" /></figure><p>This is not just an issue of wasted time. We’re also gaining an increasingly clear picture of the performance impacts of information overload.</p><p>A Harvard <a href="https://en.wikipedia.org/wiki/Knowledge_economy">research study</a> documented the decidedly non-linear effects of access to information, outlining that pretty quickly, performance doesn’t just plateau but is actually <em>reduced</em> by receiving more information. This issue manifests itself not only in less accurate recall of relevant information but also in less effective overall decision-making. As <a href="https://www.executiveforums.com/single-post/2017/03/03/Information-Overload-What-is-it-doing-to-your-employees">another study</a> put it, workers subjected to information overload “produce less output, think superficially, and generate fewer new ideas — despite working an increasing number of hours”.</p><p>In many ways our understanding of the costs of information overload at work is following the same path that the mindfulness revolution tracked nearly a decade ago. As the amount of actionable information we’re expected to process continues to increase, and evidence of the costs of information overload continues to mount, workplaces can no longer ignore the transformative benefits of predictive technology as a solution.</p><p>What makes predictive technology so powerful is its ability to optimize the information burden on humans. Why is Google Maps so popular? Because in calculating your route to work, it doesn’t just measure the shortest distance, it also analyzes historical data like what the historical traffic was like — yesterday, last week, a year ago — while also taking into account information like daily calendar events and adjusts for whether it’s a normal workday, a holiday, or even if there is road race taking place that will cause street closures. This was all information that humans used to (or forgot to!) have to process. The information burden is now reduced to remembering your car keys.</p><h3><strong><em>2. AI has transformed our predictive capabilities</em></strong></h3><p>The second force propelling us towards work becoming more predictive is the simple fact that predictions are now so much better, powered by artificial intelligence.</p><p>Effective predictions rely fundamentally on the ability to <a href="https://www.datasciencecentral.com/profiles/blogs/5-ways-to-improve-the-model-accuracy-of-machine-learning">find meaningful patterns within data sets</a>. As <a href="https://medium.com/@db_57288/ai-is-about-to-change-the-way-we-all-work-just-not-in-the-way-you-think-b9e0a52a8227">I’ve written previously</a>, this is a task for which AI is astonishingly well suited. AI’s unparalleled capacity to handle more inputs and generate better insights have transformed our predictive capacity.</p><h4><em>More inputs from more data and better sources</em></h4><p>The ability to analyze more data is important — AI’s pioneering <a href="https://www.forbes.com/sites/forbestechcouncil/2017/12/01/for-machine-learning-its-all-about-gpus/#2cee362f7699">use of GPUs has increased throughput computing capacity</a> by upwards of 20x. However, the truly transformative benefit of AI comes from its ability to add additional data sources in the form of unstructured data to our analysis. Unstructured data refers to things like text, voice, audio, and video, as well as other types of more complex overlays such as sentiment analysis, to the traditional use of structured data like numbers and dates or other pre-determined input forms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/938/1*1V3qZl8YUpmxTMagHblS4g.png" /></figure><p>The addition of these rich and varied unstructured data sources is what pushes us meaningfully past traditional analytics, and gets us much closer to approximating human-level insight. And when combined with greater volume, we’re suddenly dealing with human-level processing but at exponentially greater scale.</p><h4><em>Better insights from deep learning</em></h4><p>Having human-level capabilities at exponential scale is certainly impressive, but even beyond that, AI has transformed our prediction accuracy through the application of machine learning methods like deep learning.</p><p><a href="https://medium.com/@db_57288/three-truths-to-help-you-get-smart-fast-about-ai-today-fac4f9a078f">Deep learning</a> refers to the process by which computers use multi-layered (‘deep’) algorithms to mimic brain functionality. What makes this approach so effective for predictive purposes is the ability of these models to perform what is known as <a href="https://machinelearningmastery.com/what-is-deep-learning/">automatic feature extraction</a> — which is the ability to identify complex functions from its own mapping process. And this is the innovation that takes us beyond a reliance on human direction. This is where the models start to ‘think’ for themselves by finding relevant patterns and insights that are simply not observable by human analytic capabilities.</p><p>As one as one <a href="https://www.forbes.com/sites/forbestechcouncil/2017/06/28/the-machine-learning-imperative/#5987591e55d7">Silicon Valley technology consultant</a> puts it, when it comes to predictive technology, AI is not only better at crunching the numbers than humans will ever be but it also “allows us to answer the questions we didn’t even know we should ask.”</p><h3><strong>3. The consumerization of enterprise technology is complete</strong></h3><p>The final reason that 2019 promises to be the year that you start to see predictive capabilities come to the workplace is the one that is most overdue: the consumerization of enterprise technology is now complete.</p><p>Over the past 20 years — driven largely by IT and procurement departments overzealously justifying their existence — enterprise technology had become a dreary backwater of endless feature lists and almost now thought of user experience.</p><p>Thankfully the ubiquity of consumer technology, driven in large part by the iPhone, has refactored consumer expectations around technology portability and performance, while the rapid spread of cloud computing has improved accessibility for enterprise consumers by removing the need for large-scale software installations.</p><p>To get a sense of this change, just look at the shift in the ratio of designers to engineers hired by enterprise software companies over the past. In just five years it went from around <a href="https://www.kleinerperkins.com/perspectives/internet-trends-report-2017">1:50 to now 1:7</a>. For perhaps the ultimate evidence fo this change, no further than the fact that in 2017 one of Fast Company’s <a href="https://www.fastcompany.com/3068431/the-10-most-innovative-companies-in-design-2017">10 Most Innovative Design Companies</a> was a small business tax preparation app.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wL5onpRUgldPU_nD_DKwhw.png" /><figcaption>Source: Kleiner Perkins Internet Trends report, 2017 p.188</figcaption></figure><p>Consumers now demand that the technology at work feels like the technology in the rest of their lives. As the legendary Swiss designer, <a href="https://en.wikipedia.org/wiki/Yves_B%C3%A9har">Yves Bahar</a> pointed out: “design accelerates the adoption of new ideas.”</p><p>But the importance of consumerization is not just about better design principles, it represents is the fundamental change in enterprise technology from something that is tolerated (i.e. mandated top-down), to something that is desired (i.e. generating bottom-up demand). This is the force that will drive the biggest acceleration in the adoption of enterprise technology. Increasingly, you will be able to adopt the enterprise technology you want, when you want it, independent of your organization’s IT mandate.</p><p>This reality is what has made the fastest growing enterprise technology ever. Slack applied simple social messaging principles like group chats, emojis, gifs and memes to reinvent the category of workplace communications. But what made it truly transformative was the way it drove adoption from users up through businesses, rather than the other way around. It didn’t need to convince business leaders it was better, it simply pointed to the hundreds of existing users within each business to sell enterprise solutions.</p><p>This is the model that will take hold in 2019 and which will be the driving force in the accelerated adoption of enterprise technology.</p><h3><strong>What does this mean for you in 2019?</strong></h3><p>Each of these three trends — the information overload crisis, the application of AI, and the consumerization of enterprise technology — suggests that 2019 is going to be the year that work finally gets predictive.</p><p>Practically speaking there is much you can do to be an early beneficiary of this.</p><p>Already, there are numerous AI-powered predictive applications in-market you can use today to make your work easier, more efficient and more productive — from operational improvements like predictively managing scheduling with x.ai to functional improvements like predictively optimizing your revenue generating funnels with Lattice Engines. Chances are that if there’s an area of your work that is inefficient, there is a predictive technology solution worth trying out.</p><p>And others’ experiences have suggested that it’s worth giving it a shot. In a <a href="http://blog.speedeondata.com/articles/the-rise-of-predictive-analytics">study</a> of the performance of early predictive analytics introductions, 90% of respondents reported positive ROI on technology that had been introduced into their organization, and over 50% said that even their least successful deployment still resulted in a net benefit to their work.</p><p>And so if these trends continue, and if the effect of this technology on our work is anything like what we’ve seen in the rest of our lives, we have very good reason to look forward to 2019 looks like being the most impactful year in enterprise technology that we’ve seen in a generation.</p><p>At Asta, we’re applying predictive technology to transform the flow of information at work. You can check it out <a href="https://chrome.google.com/webstore/detail/asta/gleedeafjlflbnjkopdccnnjaegicfgh/">here</a>.</p><p>And you can also <strong>follow</strong> <strong>me</strong> directly on <a href="https://www.linkedin.com/feed/">LinkedIn</a> and <a href="https://twitter.com/danielmhburke">Twitter</a> for more on our AI future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c092414ac8ee" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/this-technology-will-make-2019-the-most-impactful-year-for-work-in-a-generation-c092414ac8ee">This technology will make 2019 the most impactful year for work in a generation</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI and the Ballad of Automation]]></title>
            <link>https://medium.datadriveninvestor.com/ai-and-the-ballad-of-automation-b1679d6f46fd?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/b1679d6f46fd</guid>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[enterprise-software]]></category>
            <category><![CDATA[software]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Tue, 31 Jul 2018 13:42:06 GMT</pubDate>
            <atom:updated>2018-08-01T20:02:31.401Z</atom:updated>
            <content:encoded><![CDATA[<h4>What tech reporters get wrong about AI, and why Systems of Intelligence are about to eat the world</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/698/1*EDswA_UDBwPWaBY22HV0zA.jpeg" /></figure><p>Perhaps the fastest-growing genre of tech news today goes like this: <em>‘AI performance disappoints, experts predict chaos.’</em></p><p>No doubt this makes pretty good journalistic fodder but articles like this don’t do much for understanding the current state or likely future value of AI technology. The foundational mistake they all make is to conflate automation and augmentation when it comes to AI.</p><p>The core value of AI lies in its ability to <em>augment</em> human capabilities. And this reality is about to power an entirely new wave of value from technological development.</p><p>To help understand why, we’ve drawn from our experiences in building enterprise AI technology at ViewX to outline why automation is so hard, why augmentation is so powerful, and why ‘System of Intelligence’ is the next tech phrase you’re about to hear lots more.</p><h3><strong>1. Automation is hard, and is not getting much easier</strong></h3><p>The single most persistent reality to appreciate about AI is that human capabilities remain extraordinarily difficult to replicate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7g68M6z_3ZgKnWdIcC7aLg.jpeg" /><figcaption>NASA’s Valkyrie robot. Just don’t give it a bottle</figcaption></figure><p>Consider robotics as an example: A <a href="https://www.therobotreport.com/10-biggest-challenges-in-robotics/">study</a> released this year by the influential journal <em>Science Robotics</em> noted that in the pursuit of robotics technology, “the major challenges… have remained largely unchanged for 30 years”.</p><p>Even NASA’s $2M <a href="https://www.inverse.com/article/30787-robotics-problems-hands">Valkyrie humanoid robot</a> — built to accomplish tasks in extreme space environments — still struggles with the kind of hand dexterity that you’d expect from most human toddlers.</p><p>If these challenges exist with respect to hand movement, consider the difficulty of replicating the performance of the human brain.</p><p>The human brain is a miraculous mess of billions of neurons that come together to operate at an estimated 1 exaFLOP, or one billion billion calculations per second. The <a href="https://www.nytimes.com/2018/06/08/technology/supercomputer-china-us.html">world’s fastest supercomputer</a>, by comparison, has a processing power of about 1/50th of that.</p><p>Indeed, in 2014, a group of researchers attempted to apply their supercomputer to process one second of one percent of the type of work the human brain is able to do. It took it 40 minutes to complete.</p><p>The reason that the human brain is so difficult to replicate is that it is enormously flexible, powerful and efficient. Its decision-making capabilities are based not just on intelligence, but also on our <a href="http://time.com/4960778/computers-smarter-than-humans/">“massively parallel processing wetware”</a> which consists of things like human instincts, common sense, and experiences. Computers can be programmed with vast libraries of information, but — without getting too poetic about it — will simply never be able to live life as humans do.</p><h3><strong>2. Augmentation is far easier, and it’s getting better every day</strong></h3><p>Whilst the brain is miraculous, it is also has major limitations. And so, while replicating human capabilities is difficult, augmenting them is not.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-JTue8sAu6pBSTcmLyxx5A.jpeg" /><figcaption>Computers are no brain. But brains are also no computers</figcaption></figure><p>What’s more, computers possess enormous capability advantages over humans in multiple broad areas: they have vastly superior memories, they can find and recall information much faster, they can be taught quicker, they are more reliably accurate in processing complex information, and they are overall less influenced by external factors.</p><p>The vast improvements in computing power over the past few decades have transformed our understanding of computational capabilities. Fortunately, these advances have come at the perfect time.</p><h3><strong>3. As information gets more valuable, computing capabilities become more critical</strong></h3><p>As you’ve probably noticed, we live in an information economy, one in which success is increasingly dependent on the quantity, quality and accessibility of information available to us.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/605/1*IiMQLziXEy7FLBlPmJYOOw.png" /><figcaption>The Economist, featuring an unusually pun-free title</figcaption></figure><p>Indeed, information is now so critical to economic success that that <a href="https://www.economist.com/leaders/2017/05/06/the-worlds-most-valuable-resource-is-no-longer-oil-but-data">The Economist</a> recently declared data — replacing oil — as the world’s most valuable resource.</p><p>Yet data and oil differ in many respects. The most important of which is scarcity. Today’s data is not scarce. The combination of the internet and the digitization of the economy combines to produce an astonishing <a href="https://www.domo.com/learn/data-never-sleeps-5?aid=ogsm072517_1&amp;sf100871281=1">2.5 quintillion bytes of digital information</a> every single day. Moreover, this information is expanding at such a rate that every two years we roughly double the amount of data that has <em>ever </em>existed in all of history before it.</p><p>What this means is that we’ve reached an information tipping point, where quantity of data becomes less of a differentiator, and can become more of a disadvantage. <a href="https://becominghuman.ai/training-data-is-the-new-oil-where-should-i-drill-284229f0dae9">Unless it’s properly harnessed</a>.</p><h3>4. <strong>AI is really good at synthesizing information</strong></h3><p>This is where the unique value of AI starts to come into focus. And especially <a href="https://www.technologyreview.com/s/513696/deep-learning/">deep learning</a>.</p><p>Deep learning thrives on large datasets. This is not only because large datasets provide sufficient training data to see algorithm improvement start to occur, but also because large datasets provide sufficient complexity for the capabilities of deep learning algorithms to really shine through.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/524/1*adq1FAG6haCzKZ9nqwBguA.png" /><figcaption>Deep Neural Networks: the Energizer Bunny of algorithms</figcaption></figure><p>Traditional algorithm performance tends to plateau at certain database sizes, limiting their impact. On the other hand, deep learning performance continues to improve as data inputs increase. What this means is that with sufficient data, deep learning algorithms continue to generate increasingly sophisticated insights and patterns, leaving conventional alternatives further and further behind.</p><p>This gives deep learning enormous power. And at ViewX we’ve seen the extraordinary benefits of this capability first-hand. Using deep learning to power our proprietary contextual recommendation technology, we’re able to process vast amounts of unstructured data to identify the most contextually relevant connections between various sets of information.</p><p>Moreover, the more data we feed it, the more performance has continued to improve, to the extent that having now processed over 100M human interactions, we’re now seeing our AI-powered recommendations outperform human selections by up to 155%. The usefulness of this is that we’re able to powerfully process huge amounts of information in a way that delivers our human users the specific and highly-relevant information they need to improve their decision-making and increase their productivity.</p><h3><strong>5. The rise of systems of intelligence</strong></h3><p>This change in our relationship to data marks the beginning of a new phase within the information age.</p><p>Until now, the winners have been those that have generated vast amounts of valuable information, yet this will no longer be enough.</p><p>The winners of this next era will be defined by their ability to not just generate data, but make use of it, particularly in a way that augments the unique decision-making capabilities of their users.</p><p>This reality is succinctly captured by the concept of a System of Intelligence, described by <a href="https://medium.com/u/cbf936487bf4">Jerry Chen</a> at Greylock Partners as <a href="https://news.greylock.com/the-new-moats-53f61aeac2d9">“the fountain of the next generation of great software companies”</a>, and which refers to any application that is able to apply AI to connect vast data sets, business processes, and user workflows, to generate valuable predictive insights for their users, whether they be customers or employees.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PHcWRBGlsVt87SPkw1Ikeg.jpeg" /><figcaption>The worker of the future: a human, but better</figcaption></figure><p>This next wave of technology will be defined by AI that augments human capabilities, rather than tries to automate them.</p><p>We’re excited to be at the forefront of this change, applying deep neural networks to build the first true system of intelligence for the fastest-growing facet of human capital management: workforce development.</p><p>The reality facing every workforce today is that one in two workers complain that they don’t have the information they need to succeed. When information is critical to success, it’s no surprise that low-information workers are less productive, less engaged, and more likely to be looking for other work. This situation currently describes 50% of your workforce.</p><p>AI provides a transformative solution, and we’ve harnessed it to build a product that understands what every worker is working on in real-time and predictively delivers them the information they need to do their job.</p><p>By using AI to personalize the flow of information to every employee, we’re augmenting their innate human ability to add context and make effective decisions. In doing so, we’re building an AI-enabled workforce that is more informed, more engaged, and more productive than ever before.</p><h3>What this means for the AI future</h3><p>What we’ve learned from building enterprise AI technology is that the value of AI lies in its ability to augment human capabilities rather than automate them.</p><p>And though we don’t expect to see the end of the media’s fascination with ‘AI performance’ stories any time soon, we do expect to see a huge rise in companies building the ‘systems of intelligence’ that use AI to transform human capabilities and power the next wave of the technological revolution. We’re excited to be riding the wave.</p><p>Interested in hearing more about how ViewX is building the <strong>AI-enabled workforce</strong>? Visit us <a href="https://www.viewx.co/#new-page">here</a></p><p>You can also <strong>follow</strong> <strong>me</strong> directly on <a href="https://www.linkedin.com/feed/">LinkedIn</a> and <a href="https://twitter.com/danielmhburke">Twitter</a> for more on our AI future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b1679d6f46fd" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/ai-and-the-ballad-of-automation-b1679d6f46fd">AI and the Ballad of Automation</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Three truths to help you get smart, fast about AI]]></title>
            <link>https://medium.com/@db_57288/three-truths-to-help-you-get-smart-fast-about-ai-today-fac4f9a078f?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/fac4f9a078f</guid>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[enterprise-technology]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[enterpreneurship]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Tue, 03 Jul 2018 02:22:20 GMT</pubDate>
            <atom:updated>2018-07-03T21:53:08.028Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*KD-62Iy5FrYRKPsCDHGW5A.png" /></figure><p>Every day we’re seeing exciting new examples of AI’s impact on industry.</p><p>As the reality of an AI-powered future becomes clearer, more and more people are reaching out for help in wrapping their heads around the application of AI to their operations.</p><p>This is the right time for these questions. As I’ve written <a href="https://medium.com/@db_57288/if-ai-is-the-new-electricity-these-three-principles-can-help-your-business-avoid-ending-up-the-c7ffb2091a17"><strong>before</strong></a>, enterprises that aren’t already actively incorporating AI into their operations risk ending up the steam powered factories of the 21st century.</p><p>Yet it’s clear that AI remains a mystifying topic for many. It doesn’t have to be.</p><p>Based on our work applying AI to the enterprise, we’ve distilled three core truths to help you get smart, fast about AI today.</p><h3>Truth 1: AI is an enabler of human intelligence, not a replacement for it</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/756/1*1Vd8gqtiCUYeEDxnye78qQ.png" /><figcaption>If they were really smart, they’d team up</figcaption></figure><p>By far the greatest source of confusion around AI is the difference between artificial and human intelligence.</p><p>Whilst recent advances in <a href="https://medium.freecodecamp.org/want-to-know-how-deep-learning-works-heres-a-quick-guide-for-everyone-1aedeca88076">deep learning</a> — the process by which computers use multi-layered (‘deep’) algorithms to generate predictive insights — has expanded the capabilities of computers, the type of intelligence that humans take for granted — creativity, context, meaning and commonsense reasoning — remain firmly out of reach.</p><p>Instead, the reality of AI is that it is astonishingly powerful at a pretty narrow range of activities. These mostly revolve around identifying patterns within data sets. That’s not to say it isn’t extremely valuable, it is. It’s just not the same as human intelligence.</p><p>Indeed its value comes from its complementarity to human intelligence. AI works best when its computational processing power is used to inform human decision-making.</p><p>We see this in all forms of applied AI. Even in chess, where AI has been largely superior to human ability for 20 years, the most successful players still remain <a href="https://www.huffingtonpost.com/mike-cassidy/centaur-chess-shows-power_b_6383606.html">hybrid human-AI teams</a>, which consistently mop the floor with their AI-only and human-only opponents.</p><p>The same principle applies in enterprise. At ViewX we use this truth to inform the way we’re harnessing AI to reimagine workforce development technology. Our neural networks are able to interpret task-level email data to predictively match each worker with the most relevant video information in that moment.</p><p>In other words, rather than trying to replace human intelligence, we’re using AI to deliver highly relevant information to human workers in a way that enhances human intelligence. AI reinagines our ability to process information, improving decisions, productivity and performance as a result.</p><p>The first truth to remember: <strong>AI is far more powerful as an enabler of human intelligence, than a replacement for it</strong>.</p><h3>Truth 2: AI will only ever be as good as its data</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XDK_YtpsAOw0ysVUHwZL6A.png" /></figure><p>If the difference between AI and general intelligence is the most misunderstood part of AI, then the fundamental importance of data is the most under-appreciated.</p><p>The far-reaching implications of data quality in AI were recently demonstrated by a team of researchers recently.</p><p>They studied the impact of gender biases <a href="https://www.wired.com/story/machines-taught-by-photos-learn-a-sexist-view-of-women/">found</a> in two commonly-used image training data sets, both of which inadvertently reflected gender stereotypes by tending to tag women in activities like shopping and cooking, while activities like work and hunting were commonly linked to men.</p><p>The deep learning algorithms they trained on this data set began to routinely identify an image of a man standing in a kitchen as “woman”, while in another it was asked to complete the statement “Man is to computer programmer as woman is to…” and would provide responses like “homemaker”.</p><p>Worse still, the nature of deep learning means that AI trained on bad data will not merely reflect its failings, but actually amplify them.</p><p>In the stereotypes study, the frequency of gendered outputs ended up higher than the original biases in the data set because they were continually reinforced by the details of that labelled data.</p><p>This is a critical consideration for any business decision-maker, who not only needs to understand the quality of the training data used to develop the algorithms they’ll be relying on, but also needs to consider the quality of the ongoing data from their own businesses that will be used to generate AI insights.</p><p>Our Head of Product, Alex Gould, is a specialist on data design and has explored this in a great piece on the data design principles that we use for our AI products <a href="https://medium.com/@ag_6063/training-data-is-the-new-oil-where-should-i-drill-284229f0dae9">here</a>.</p><p>All this underscores the second fundamental truth: <strong>AI performance will only be as good as its input data</strong>.</p><h3>Truth 3: AI mistakes will look different to human mistakes</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*Wp0xAV28G6HvbrZYYNgHzQ.png" /></figure><p>The final truth about AI, which gets by far the least attention but which is shaping up to be among the most interesting to watch as AI adoption matures, is that AI makes very different mistakes compared to humans.</p><p>The impact of this is that human users find it difficult to empathize with AI. The concern about a lack of empathy for AI may seem like a very solid ‘nice-to-have’, but this lack of empathy is damaging because it usually manifests itself in unwarranted disillusionment with AI products in general.</p><p>This issue will become an increasingly important practical consideration for business decision-makers looking to drive AI adoption in their organizations.</p><p>It is already creating practical barriers. An example is in the growing field of AI-powered scheduling assistants, which are facing adoption challenges in part because their models make seemingly ‘stupid’ mistakes that are obvious to humans. Consider an email sent at 1AM to a group saying, “Let’s get together first thing tomorrow to finalize this, and let’s add Julie”. The AI assistant is likely to make its error from incorrectly interpreting the ambiguity in “1AM” and “tomorrow”, which most humans would be able to navigate pretty easily. On the other hand, the error of a human scheduler is more likely to be an error of oversight, forgetting to include Julie on the invite at all, for example.</p><p>This is no less of an error, but it is one that many humans have made themselves or can otherwise more easily empathize with. This empathy gap is so pronounced that researchers are starting to program ‘human-style mistakes’ <a href="http://www.businessinsider.com/humans-like-robots-that-make-mistakes-2015-10">intentionally</a> into AI programs to make them more likable.</p><p>We believe that as human users get more exposure to AI performance, this will improve. Until then this reality will require understanding and expectations-setting on the part of AI sponsors within organizations mindful of the third truth: <strong>AI mistakes will look different to human mistakes</strong>.</p><h3>Lessons to take from these three truths</h3><p>These three truths are designed to provide a simple starting point for anyone looking to get smart, fast on the basics of AI.</p><p>Taking them one level further, they can also provide a useful guide for business owners looking to plan for AI adoption.</p><p>Understanding AI’s value as an enabler of human intelligence helps you define the right role for AI in your organization.</p><p>Appreciating the importance of data helps you focus on the right details needed for optimal AI performance.</p><p>And being aware of the differences in AI and human mistakes helps you pre-empt the empathy challenges of human users.</p><p>In all, AI can be a daunting topic, but it doesn’t have to be. We hope that, armed with a better understanding of these AI truths, you’ll now be in a better position to start planning effectively for the coming impact of AI on your industry.</p><p>Interested in hearing more about how ViewX is building the <strong>AI-enabled workforce</strong>? Visit us <a href="https://www.viewx.co/#new-page">here</a></p><p>You can also <strong>follow</strong> <strong>me</strong> directly on <a href="https://www.linkedin.com/feed/">LinkedIn</a> and <a href="https://twitter.com/danielmhburke">Twitter</a> for more on our AI future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fac4f9a078f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI is about to change the way we all work, just not in the way you’ve been hearing]]></title>
            <link>https://medium.com/@db_57288/ai-is-about-to-change-the-way-we-all-work-just-not-in-the-way-you-think-b9e0a52a8227?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/b9e0a52a8227</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[employee-engagement]]></category>
            <category><![CDATA[human-resources]]></category>
            <category><![CDATA[enterprise]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Wed, 06 Jun 2018 13:08:46 GMT</pubDate>
            <atom:updated>2018-08-27T01:14:29.597Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vhNlfHxIyfXfzgF8DzzGNw.jpeg" /></figure><p>Much of the discussion today about AI and the future of work focuses on the future threat of automation and job losses.</p><p>This is an important discussion, but it remains premature. Even the rosiest estimates put meaningful automation at two to three decades away. As Elon Musk recently tweeted after his Tesla factory was forced to shut down in part due to an over-reliance on automation: “humans are underrated”.</p><p>Indeed, there’s a far more pressing crisis hitting these underrated human workers: an astonishing two out of every three workers say they’re ‘disengaged’ at work, according to a recent Gallup poll.</p><p>This is a crisis that touches every workplace. Engagement and innovation are inextricably linked, and our failure to address this costs the economy an estimated $500B in lost output every year.</p><p>And solutions have been hard to come by.</p><p>Our work applying AI in enterprise has given us three insights into this crisis and the ability for AI to offer the capabilities necessary to solve it. What we’re seeing is that in helping to finally solve it, AI is beginning to look far more likely to be a fundamental enabler of future human workforces than a threat.</p><h4><strong>1. The cost of the workforce engagement crisis</strong></h4><p>The fact that two in three workers are disengaged at work is a chilling statistic for three reasons. First, low-engagement employees are expensive: they’re roughly 20% less productive and 90% more likely to churn than their more engaged co-workers. Research suggests that the cost of employee turnover alone can run between 30%-100% of a churned employee’s annual salary.</p><p>Second, engagement and innovation are inextricably linked. A recent survey mapped employees’ engagement with their performance and found that those who were most engaged were up to 20x more likely than their disengaged colleagues to go out of their way to find innovative solutions to challenging customer problems.</p><p>Third, as we continue to transition to a knowledge economy — where growth depends more on effective use of information than the means of production — the value of human innovation is only going to increase. As computers continue to take away more and more menial tasks, nearly every workplace is seeing the estimated 60% shift in the proportion of work that relies primarily on capabilities unique to the human brain like critical thinking and making complex judgments.</p><p>The reality is that an engaged workforce is an increasingly important source of competitive advantage, yet businesses are struggling more than ever to maintain this.</p><p><strong>Insight 1: </strong>Two-thirds of today’s employees are disengaged at work. This affects every business in adding unnecessary operating costs. More importantly, it undercuts an increasingly important source of future competitive advantage for businesses in a knowledge economy: human innovation.</p><h4><strong>2. Work has changed. Workforce enablement technology has not</strong></h4><p>The cause of this engagement crisis can be traced in large part to the fact that the nature of work and our workforce have changed, while the way we inform, engage and equip these workers has not.</p><p>Over the past decade, work has shifted from career-based to task-based employment, a change which has made work more transactional, transient and remote than ever before. Workers are 3x more likely to have started a new job in the past year, 50% will work remotely some or all of the time, and 93% of employers are actively pursuing flatter management structures.This change has benefited millions in the form of better access to more lucrative opportunities, but it has also come with a cost. Task-based work trades convenience for connectedness, while transience and technological change mean that workers are required to learn more, more often and with less institutional support than ever before.</p><p>Despite this change, the way we inform, engage and train our workers has barely evolved. Fully 60% of Fortune 500 companies have no dedicated workforce development system. And those that do probably shouldn’t have bothered: existing corporate development programs receive a score of -31 NPS from their users. Workers are still presented with the usual suspects of the past: stale corporate intranets, outdated learning portals, and siloed information sources, so it’s little wonder that today’s workers only report spending 1% of their work week attending to their own learning and development. This is an unsustainable situation given the demands of task-based employment.</p><p>Demographic shifts are about to make this problem worse. Just as Millennials changed the culture of work, the impending entry of Gen Z — the first digital native generation — into the workforce promises to transform the technology of how we work. This is a generation that has grown up on a highly personalized digital diet of continual engagement and constant content consumption. Seventy percent of this group check their smartphones 30+ times an hour and 60% watch more than two hours of digital video each day. With 75% of the workforce forecast to be a combination of millennials and Gen Z in just five years, we are entering the era of the Snapchat generation at work. The company intranet is no longer going to cut it.</p><p><strong>Insight 2: </strong>The shift to task-based work has fundamentally changed the development needs of our workforce. Yet the way we inform, engage and train our workers remains stuck in the static systems of the past. These static solutions are manifestly unable to address our workforce engagement crisis. Things are about to get worse with the impending entry of the first digital native generation into the workforce.</p><h4><strong>3. AI provides a next-gen solution for a next-gen workforce</strong></h4><p>Meeting the development needs of a digitally native workforce engaged in task-based employment will require a paradigm shift in the way we inform, engage and train our workers.</p><p>AI looks increasingly likely to provide the technological path to enable this solution.</p><p>The core capability of AI is its ability read, analyze and understand large sets of unstructured data — text, video, and images — to find patterns, connections, and matches between disparate data sources.</p><p>Applied to the needs of modern workforces, it gives us the task-level ability to understand the needs of each worker and automatically match them to the right information — whether that’s learning and development, news about their industry, or content created by their company — exactly when they’re doing the specific task that requires it. This kind of context is what’s been missing for employers in informing, training and engaging their workforce.</p><p>This is the exciting prospect at the heart of the emerging $60B AI-powered workforce enablement industry. AI finally gives us the ability to continuously connect every worker with an always-on feed of highly relevant information that is hyper-personalized to their individual needs. When this technology is integrated into workflows it offers the ability to analyze the specific work done by each employee and match them to the exact information they need in real-time. Employers that can give their employees better access to highly relevant information will not only see higher engagement and lower churn, but also a workforce that is more informed and better equipped to succeed in a knowledge economy</p><p><strong>Insight 3: </strong>Workplace success depends more than ever on faster access to better information. AI finally gives us the capabilities we’ve been looking for, getting us back up to speed with the needs of today’s workers. The unique ability of AI to process vast amounts of unstructured data to deliver valuable information is the paradigm shift in workforce enablement we’ve have been looking for. Applied to areas like learning and development, internal company communications and industry intelligence, it is giving us reimagined capabilities to connect our workforces with the information they need to succeed at work.</p><h4><strong>The path forward</strong></h4><p>Every business is realizing that maintaining a competitive advantage increasingly depends on the ability to maintain an informed, engaged and trained human workforce. This requires an ability to provide workers with better access to highly relevant information whenever they need it.</p><p>AI’s core capability of processing large sets of unstructured data presents an exciting opportunity to reimagine the way we connect human workers to the information they need to succeed at work. This is why the work we’re doing in applying AI in the enterprise has convinced us that for the foreseeable future, AI is far more likely to be a fundamental enabler of human workforces than a threat.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b9e0a52a8227" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[If AI is the new electricity, here are three principles to help your business avoid becoming the…]]></title>
            <link>https://medium.com/@db_57288/if-ai-is-the-new-electricity-these-three-principles-can-help-your-business-avoid-ending-up-the-c7ffb2091a17?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/c7ffb2091a17</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[enterprise]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Tue, 03 Apr 2018 13:41:58 GMT</pubDate>
            <atom:updated>2020-08-06T13:54:09.035Z</atom:updated>
            <content:encoded><![CDATA[<h3>If AI is the new electricity, here are three principles to help your business avoid becoming the steam-powered factory of the 21st Century</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/962/1*u8FvizJCIaNhe90G4PQ-uw.png" /></figure><p>‘AI is the new electricity’. So says <a href="https://www.youtube.com/watch?v=21EiKfQYZXc">Andrew Ng</a>, the Stanford professor and former head of Google AI, who has spent his career at the forefront of the artificial intelligence and deep learning revolutions.</p><p>After a flurry of media attention over the past 18 months, it seems this message is breaking through, with <a href="https://sloanreview.mit.edu/projects/reshaping-business-with-artificial-intelligence/">85% of business leaders</a> recently responding that AI is likely to transform their industry.</p><p>This increased attention is important, though it’s yet to translate into clear action. Only 5% of respondents in the same survey said that they’re actually utilizing AI as a meaningful part of their current operations.</p><p>This lag should worry these businesses. If AI is the new electricity, history illustrates the stark danger of inaction.</p><p>At ViewX we’ve observed two contributing causes for this: first, senior executives still think in future tense, not realizing the extent to which the AI revolution is already upon them. And second, there is almost no practical guidance for business leaders on the simple steps they should be taking now to set themselves up for success.</p><p>There are two points that every business leader needs to keep in mind now, or else risk ending up a steam powered factory of the 21st century.</p><h4><strong>Point One: It’s the year 1900 for AI</strong></h4><p>To appreciate the urgency of AI adoption it’s important to put electrification in its full context; the ubiquity of electricity today makes it easy to forget what a seismic event it had on the economy 100 years ago.</p><p>The introduction of electricity to the economy produced famous leaps in the productive capacity of businesses in virtually every industry. It produced an explosion in business capabilities, with average productivity growth rates rising to roughly five times greater than what we’re used to today.</p><p>While it produced enormous winners, it also produced huge classes of losers from those businesses who failed to act. An under-appreciated fact about electrification was that it drove the highest rate of business failure for established entities for any period in the 20th century, outside the Great Depression. That’s because it created competitive advantages for those who adopted it that effectively wiped out their steam-powered competition.</p><p>And this process happened quickly. In 1900 less than 5% of all industrial power in the US came from electricity. Steam power still reigned supreme with its reliable but limited single steam engine. Fast forward just 10 years though, and electricity’s contribution will have jumped to over 50%. Ten years after that and the electrification of the economy was complete. In just 20 years the era of steam power was over.</p><p>The relevance today of this today is that AI is the same age now as electricity was in 1900.</p><p>For electricity, the inflection point came from the construction of centralized high-voltage power stations in 1889, producing a reliable supply of electricity to business. The equivalent inflection for AI happened in the late 2000s with the discovery that GPU chips — originally designed for video games — were uniquely suited to the complex processing necessary to accurately mimic human brain performance.</p><p>This is what’s known as AI’s deep learning breakthrough; a development that suddenly opened up the massive digital data sets, that had been produced by the rapid spread of the internet and the inexorable digitization of the economy, necessary to feed the deep learning processes and transform their effectiveness. This combined growth of computing power and data supply transformed the capacity of AI to perform, and learn from, increasingly sophisticated tasks. This is why you’ve likely noticed an uptick in news stories over the past several years where AI has started to outperform humans in an increasingly wide range of tasks.</p><p>Hence the risk for businesses. If we’re in 1900 for AI, we’re about to hit a seismic period of disruption that creates huge classes of businesses winners and losers. And this change happens fast.</p><h4><strong>Point Two: Here are three simple principles to help every business incorporate AI</strong></h4><p>The lessons of electrification provide a useful starting point for how businesses can and should respond now. We’ve combined these insights with our experiences in bringing AI to the enterprise to distill three practical principles that distinguish the thinking of the most forward-thinking executives today, and which every senior executive can adopt to incorporate the benefits of AI capabilities to their business. They are:</p><p>1. Think big. Don’t settle for merely swapping out your steam-powered engine</p><p>2. Your data pipeline is your assembly line</p><p>3. Hire talent to design your wiring, not build your own grid</p><p>Adopting these principles is the meaningful first step most executives are missing, towards giving your business the best chance of thriving in this AI age, and avoiding being the steam-powered factory of the 21st century.</p><p><strong><em>1. Think big. Don’t settle for merely swapping out your steam-powered engine</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/1*65vR6F0_5hDmc3RVXliE8w.png" /><figcaption>An early steam turbine. The attachment to them is not totally clear now</figcaption></figure><p>By far the biggest failure of business owners in the 1900s was the failure to think big about the transformative potential of electricity. We see the same trap befalling uncertain executives today who tend to think of AI primarily in terms of basic automation and cost reduction opportunities. Just as with electricity, the big winners of the AI age will be those that direct their AI efforts at growing revenue and market share, not just shaving incremental cost.</p><p>For steam-powered businesses of 1900, the most common early adoption of electricity was simply to replace your steam-powered engine with an electrical one, leaving everything else about your factory intact. Not only were the results financially underwhelming — most usually failed to pay back the initial investment costs — but worse, the disappointing initial results often led early experimenters to dismiss electricity as a mere technological gimmick which had little practical business benefit and opt instead to stick with tried-and-tested steam.</p><p>As we know now, the benefit of electricity was not in its ability to produce power marginally cheaper than steam, but to completely re-define where and when power could be delivered to your workforce.</p><p>This was the insight of innovators like Henry Ford, who recognized the transformative potential of electricity to completely re-configure his factory floor in pursuit of improved productive output. Rather than seeking to do the same thing only slightly cheaper, Ford completely reimagined the production process, redesigning his factory from top to bottom so that his workers — no longer dependent on centralized power — were able to employ their tools whenever and however was best. The results of this re-thinking are now legendary, with Ford’s electrified factory seeing a 70% improvement in the productive output of his workforce, a capability that drove a five-times increase in his market share over the next six years, from 9% in 1908 to 48% in 1914.</p><p>This is the same kind of thinking that will win in AI today. Just as electricity untethered productive capacity from the limits of mechanical power, AI now allows us to extend this even further, pushing us past human limitations. It allows us to process and use information at a scale and level of complexity that — just like the ability of electricity to deliver flexible mechanical power — were previously unimaginable.</p><p>At ViewX we’ve seen a great example of this in the way our media industry customers are using our technology to transform their video capabilities. Using our AI models — which are able to automatically read and understand vast video libraries at 70x human speed and 6x human quality — have helped these companies turn video from a pain point into a critical differentiator, transforming the scope and quality of their content experiences and delivering performance improvements of up to 180% as a result of higher utilization, consumption and monetization of this video content.</p><p>The lesson to take from these applications is that if you’re thinking of AI only in terms of cost reduction, and not at top-line growth, you’re not thinking big enough; you’re basically fitting out a slightly cheaper steam-powered factory.</p><p><strong><em>2. Your data pipeline is today’s assembly line</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/568/1*mwUrMHGysIOgwZi4cfoswg.png" /><figcaption>The Model T and the model for how you should think about your data</figcaption></figure><p>The second principle today’s businesses can take from electrification is that the critical importance of reconfiguring your processes to truly activate the benefits of the technology. In 1900s, this meant innovating your assembly line. With AI it means reimagining your data pipeline.</p><p>To understand this let’s jump back to 1900 and the steam-powered factory, which would have been an awe-inspiring sight: entire floors of workstations connected by a complex network of leather belts and pulleys, all connected to a monolithic single drive shaft powered by a giant single steam engine.</p><p>However, its main strength — a reliable central power source — was also its critical limitation, requiring that the entire design of factories be based on those machines that needed to be closest to the drive shaft to get enough mechanical power. Electricity changed all this, enabling processes to be designed with production efficiency in mind, rather than merely what was needed for power supply. This was the birth of the modern assembly line. AI does the same thing today, but for data.</p><p>At its core, AI transforms the ability of today’s business to make use of information, redefining the capacity to gather, interpret and utilize masses of available data, to find previously unidentifiable connections, and to deliver insights, predictions and recommendations that improve business decision-making and transform business performance. Much like the assembly line changes during the 1900s, the business executives who are distinguishing themselves with AI today are the ones actively re-designing their business processes to collect more, and higher quality, data at every instance.</p><p>Again, we’re seeing great opportunities at ViewX in workforce enablement applications, where our AI can be used to analyze the millions of communication data points generated by their employees in day-to-day interactions to identify the skills and knowledge gaps that can be filled by delivering highly relevant information and guidance right to their fingertips. To do this requires concerted effort to identify, collect and design data capturing processes to make the most of this AI-enabled information, but once you do, research indicates that the payback is improvements worth up to 15% higher annual revenue performance from this new information-enhanced workforce.</p><p><strong><em>3. Hire talent to design your wiring, not build your own grid</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mdbSVwO8gcN-pdJ3BOc_iQ.jpeg" /><figcaption>Early on-site electricity generators. You don’t need the AI version of these</figcaption></figure><p>The third and final lesson for today’s business owners is that the winners of electrification were those that most efficiently applied electrical power, not those that tried to build their own.</p><p>The knee-jerk response of many businesses of the 1900s was to build on-site generators, mistakenly identifying production as the source of competitive advantage for electricity. It seems ridiculous now, but we see this same dynamic playing out today in too many unsure executives who assume that introducing AI capabilities requires the recruitment of internal AI functions to build their own technology. Not only does this rarely pay back, it even more damagingly encourages the belief that, if these kinds of investments are out of reach for your business, so too are the benefits of AI. Nothing could be further from the truth.</p><p>Rather, the target in-house AI capabilities for almost every company we come across is met simply by having dedicated senior executive responsibility for identifying the capabilities and processes to be targeted for AI-enhanced performance improvements. At its core, this role requires an understanding of the AI landscape, awareness of the technological capabilities, and the authority and support to identify and recommend the most impactful application of AI to your business performance.</p><p>At ViewX we’ve seen this role performed well by a variety of functions like CTOs, Chief Product Officers and Chief Data Officers, when they possess a strong technical understanding of AI and are given cross-functional mandates. Other companies leaning heavily into AI are creating Chief AI Officer or VP of AI roles to quickly ramp up AI activities. At the end of the day, title is less important than substance, and the right choice will depend on the specifics of your company’s unique context and existing executive structure.</p><p>Most importantly, the winners of the AI age will not be those that spend their time unthinkingly trying to replicate core technology, but instead be those that maintain the ability to plug into the latest technology evolution, and then are most decisive in applying their domain expertise to translate that technology into transformative business performance.</p><h4><strong>Practical steps for meaningful benefit from AI</strong></h4><p>All available evidence points to the fact that we’re in the year 1900 for AI, which if true, means that we’re about to see a period of wide-scale disruption as AI-enabled capabilities sweep through every industry.</p><p>At ViewX we specialize in delivering our proprietary AI in a way that transforms business capabilities. The three principles we’ve laid out provide a practical starting point designed to help your business actively configure itself to plug-in to AI’s productive potential and translate this technology into tangible business benefits. We’d love to hear from you if you have any questions or could use some advice as you think through your own applications.</p><p>As history teaches, the sooner and more effectively you act, the greater chance you give yourself to thrive on the benefits of this technology, and the more likely it is your business won’t end up as the steam-powered factory of the 21st century.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c7ffb2091a17" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How AI is helping to save the media industry]]></title>
            <link>https://medium.com/@db_57288/how-ai-is-helping-to-save-the-media-industry-f5876ef84b9b?source=rss-4a836b2efa74------2</link>
            <guid isPermaLink="false">https://medium.com/p/f5876ef84b9b</guid>
            <category><![CDATA[medium]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[business-performance]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[innovation]]></category>
            <dc:creator><![CDATA[Daniel Burke]]></dc:creator>
            <pubDate>Tue, 30 Jan 2018 18:37:06 GMT</pubDate>
            <atom:updated>2018-03-19T01:40:34.615Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ls8YpHMmbkCFb3Mdvn4-NQ.jpeg" /><figcaption>Media companies have identified a ‘shift to video’ as a critical pathway out of their digital dilemma. AI provides the capabilities necessary to deliver on that strategy</figcaption></figure><p>Few industries have been hit as hard by the technological changes of recent times as the media industry.</p><p>The same trends that have improved the lives of billions — the growth of the internet, the spread of social media, and the proliferation of smartphones — have instead disrupted the business models of every major media company, diluting their ability to sustainably fund their core operations.</p><p>Most media companies have identified a ‘shift to video’ as a critical pathway out of this digital dilemma. Digital video content is five times more engaging for consumers and four times more valuable for advertisers than text content alone.</p><p>However, despite huge investments by media companies in increased video production, these shifts to video have so far failed to deliver meaningful bottom-line results.</p><p>That’s because you need to accurately match highly relevant content to the right audience in order to drive video performance. And so although production has scaled, the necessary matching technology has not. Media companies are still reliant on legacy solutions like manual keyword tagging and human matching. This has left their vast video libraries both under-utilized and under-valued.</p><p>AI provides the solution.</p><p>Advances in deep learning technology can be applied to ‘read’ vast libraries of video to determine what’s in them, and ultimately be trained to ‘understand’ which videos are most relevant in a given context. Suddenly, media companies have a technological solution that can scan and surface highly relevant video to the right audiences at scale.</p><p>“This is a perfect example of AI’s ability to deliver direct benefit to business performance,” says Dr. Michael Barnathan, a former Google engineer and current CTO of ViewX, a start-up that’s built an AI used by multiple major media companies to help them transform their video performance.</p><blockquote>Despite huge investments by media companies in increased video production, these shifts to video have so far failed to deliver meaningful bottom-line results.</blockquote><p>ViewX uses deep learning to analyze the full scope of information contained in any premium video — combining sources such as face detection, audio transcription, scene segmentation, scene captioning, and optical character recognition (OCR) — and then applies multiple proprietary innovations to improve how that information is prioritized, tailored to reflect the needs of the media industry. The result: transformed accuracy that moves beyond traditional performance limitations to achieve up to 98% accuracy vs human ground truth.</p><p>Their technological innovations touch on both the metadata generation and their proprietary relevance engine. For example, their OCR processor captures on-screen text by analyzing patterns in videos’ morphological gradients to differentiate programmatically overlaid text, natural text in the background, and the underlying background image. A Long Short-Term Memory (LSTM) neural network then processes the segmented gradients to translate each stream to much more accurate human-readable text than previously possible.</p><p>ViewX’s relevance engine breaks new ground in NLP and vector based matching techniques to improve the accuracy of how videos are found and recommended based on their relevance to any input, whether articles, videos or audience types. Their users’ proprietary selection data is collected as feedback, aggregating multiple media organizations to drive continuous optimization of their results.</p><p>For Dr. Barnathan, the opportunity to apply AI directly to a critical industry challenge is the most validating part of ViewX’s journey: “A thriving media industry has never been more important nor more under threat. It’s exciting to be working with the biggest brands in media to harness the incredible power of AI to unlock the value of their video.”</p><p>With ViewX and the power of AI on their side, media companies can finally look forward to a new wave of technological advancement that will enhance rather than dilute their ability to generate sustainable profits and power the vital social function they perform.</p><p><em>Written by Daniel Burke and Alexander Gould</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f5876ef84b9b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>