<?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 TO THE NEW Insights on Medium]]></title>
        <description><![CDATA[Stories by TO THE NEW Insights on Medium]]></description>
        <link>https://medium.com/@marketing_87790?source=rss-0b5fddc2d7a5------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*v6uur84ZEX86wU5q</url>
            <title>Stories by TO THE NEW Insights on Medium</title>
            <link>https://medium.com/@marketing_87790?source=rss-0b5fddc2d7a5------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 05:17:21 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@marketing_87790/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[Why Your Frontend Team Is About to Get More Important, Not Less]]></title>
            <link>https://medium.com/@marketing_87790/why-your-frontend-team-is-about-to-get-more-important-not-less-be519c565a26?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/be519c565a26</guid>
            <category><![CDATA[frontend]]></category>
            <category><![CDATA[frontend-developer]]></category>
            <category><![CDATA[frontend-development]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Fri, 20 Nov 2026 09:23:10 GMT</pubDate>
            <atom:updated>2026-11-20T09:23:10.102Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Programmer and colleague in office training machine learning algorithms" src="https://cdn-images-1.medium.com/max/1024/1*88y73sdoA6OLo3GQpR1O_w.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p>If you manage frontend developers, you’re about to notice something strange.</p><p>Your team will ship faster. They’ll prototype in days instead of weeks. They’ll iterate on designs quicker, respond to feedback more immediately, and experiment more freely. And they’re going to spend less time writing code.</p><p>This isn’t because code matters less. It’s because code has stopped being the bottleneck. AI is handling the routine work. The bottleneck has moved somewhere else entirely.</p><h3>What Actually Changed</h3><p>A year ago, <a href="https://www.tothenew.com/services/digital-engineering">frontend development</a> was still about implementation. You had a design, you translated it into code. The developer’s value was in their ability to write that translation accurately and efficiently.</p><p>Today, that translation is becoming automated. AI tools like <a href="https://github.com/features/copilot">GitHub Copilot</a> can generate context-aware components from descriptions. Design tools use AI to generate layout variations and design system components. AI libraries can suggest accessibility patterns, performance improvements, and component structures automatically.</p><p>The developer no longer needs to be the person who translates designs into code. They’re becoming the person who decides whether the translation is right. That’s a completely different job.</p><h3>How Is Frontend Development Actually Changing?</h3><p>Here’s what this means in practice.</p><p>A developer used to spend 60% of their time writing boilerplate, handling accessibility concerns, optimizing performance, and generating documentation. They had maybe 40% of their attention left for thinking about architecture, edge cases, and user experience.</p><p>With <a href="https://www.tothenew.com/insights/article/what-is-ai">AI</a> handling the routine work, those numbers flip. Now they can spend 60% of their time on the decisions that actually matter. Architecture. User experience. How different features interact. What happens when systems fail.</p><p>The developer becomes a curator rather than a producer. They’re evaluating what AI generated, thinking about whether it solves the right problem, understanding the tradeoffs between different approaches.</p><p>That requires more critical thinking, not less. It requires deeper understanding of the business problem, not just the technical implementation.</p><h3>How Does AI Change Team Collaboration?</h3><p>There’s another shift happening at the same time.</p><p>When developers had to write everything from scratch, design and development were separate phases. The designer finishes, throws it over the wall, and the developer builds it. Feedback comes back. Changes take weeks.</p><p>With AI tools creating a shared language between design and development, that barrier disappears. A designer can create a variation in Figma and immediately have a working component. A developer can suggest an accessibility improvement and see it reflected in the design system instantly.</p><p>Designers and developers start working together instead of sequentially. Product managers can iterate on user flows with real feedback. Teams across different functions start thinking about the problem together rather than handing it off. The team becomes smarter because everyone has visibility into everyone else’s decisions.</p><h3>Why Does Quality Matter More When Speed Increases?</h3><p>Here’s where most organizations get it wrong.</p><p>They see <a href="https://www.tothenew.com/insights/article/ai-agents-in-frontend-development">AI speed up frontend development </a>and think they can ship more with fewer people. So they ship faster but skip the review. They deploy more but test less carefully. They celebrate the speed.</p><p>Then they ship bias because the AI was trained on biased data. Accessibility breaks on certain devices because the AI didn’t understand the edge case. Code becomes unmaintainable because the AI optimized for the wrong metric.</p><p>Speed without judgment isn’t speed. It’s technical debt accumulating in real time.</p><p>The organizations that win are the ones that flip the equation. They use AI to move fast, but they invest the time they save in deeper review. They test more rigorously. They think more carefully about edge cases and ethical implications. They maintain higher standards for code quality and accessibility. Speed gives them a window. Quality keeps them ahead.</p><h3>What Should You Do About This?</h3><p>If you’re building a frontend organization, this shift changes what you hire for and how you structure teams.</p><p>You need developers who understand architecture and can think about systems holistically. You need people who can communicate with designers and product managers about why a solution does or doesn’t work. You need developers who care about code quality and accessibility, not just features shipped.</p><p>You need fewer people who can write boilerplate and more people who can make good decisions.</p><p>The salary floor for frontend developers is about to go up. The demand for good decision-makers is going to outpace supply. The developers who can think about problems deeply and evaluate tradeoffs will be the ones organizations compete for. This is good news for the profession. Developers have always wanted to do more than write code. Now they finally can.</p><h3>What Does Your Organization Need to Do?</h3><p>This shift only works if you have the right processes.</p><p>You need code review practices that actually evaluate what AI generated. You need testing standards that catch the edge cases AI missed. You need accessibility checkpoints that prevent the bias from shipping. You need conversations with designers and product teams that happen before code is generated, not after. The tools give you speed. The process gives you confidence in that speed.</p><p>Organizations that invest in process alongside tools will ship faster and more reliably. Organizations that use tools without updating their process will ship faster and hit walls constantly.</p><h3>What’s the Real Opportunity Here?</h3><p>The real opportunity isn’t the speed. It’s what your team can do with the speed.</p><p>A team that used to spend three weeks building a feature can now spend one week building it. What do they do with the two weeks they saved? They could ship more. Or they could build better, care more about quality, support more ambitious products, and move faster on the work that actually matters.</p><p>The organizations that treat this as an efficiency play will get some benefit. The organizations that treat it as an opportunity to rethink how they build will get a lot more.</p><p>Your frontend team is about to become more valuable, not less. But only if you help them move into the work that matters and give them the tools and processes to do it well.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=be519c565a26" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Platform Engineering Shift That Separates Fast Companies From Slow Ones]]></title>
            <link>https://medium.com/@marketing_87790/the-platform-engineering-shift-that-separates-fast-companies-from-slow-ones-01de6971229d?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/01de6971229d</guid>
            <category><![CDATA[engineering]]></category>
            <category><![CDATA[platform-engineering]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Thu, 19 Nov 2026 04:52:03 GMT</pubDate>
            <atom:updated>2026-11-19T04:52:03.293Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Close up of data center programmers using pc to visualize ai neural networks" src="https://cdn-images-1.medium.com/max/1024/1*6vQjU9Wc_sj1KFC1iV9nHg.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p>A decade ago, DevOps solved a critical problem: developers and operations teams were siloed, and software delivery was glacial.</p><p>Today, in organizations that never moved past that first solution, DevOps has become the problem. The best-performing technology companies have already moved on. They’re not managing <a href="https://www.tothenew.com/blog/devops-evolving-with-ai/">DevOps </a>anymore. They’re building platforms.</p><h3>Why Does Speed Actually Slow Down as You Scale?</h3><p>Here’s what happens in most large organizations.</p><p>Fifty developers, each with different infrastructure needs. So operations creates tools and frameworks to serve them. Teams start customizing. Someone builds a wrapper around the wrapper. A few teams create their own deployment pipelines because the standard one doesn’t quite fit their use case.</p><p>A year later, you have fifty developers interacting with dozens of infrastructure systems. Each developer spends 30% of their time managing systems rather than building products. Deployment that used to take days now takes weeks, partly because of the complexity and partly because every team is following a slightly different process.</p><p>This isn’t a failure of DevOps thinking. It’s the natural consequence of scaling it. When you empower fifty teams to make infrastructure decisions independently, you get fifty different approaches. Fast companies saw this trap coming. They created a different kind of infrastructure.</p><h3>What Does Platform Engineering Actually Solve?</h3><p>A platform is not another tool. It’s a layer that sits between developers and all the complexity they would otherwise have to manage themselves.</p><p>Instead of developers <a href="https://www.tothenew.com/blog/kubernetes-deployment-using-istio-service-mesh/">interacting with Kubernetes</a>, cloud configuration, CI/CD pipelines, monitoring systems, and security controls, they interact with a single interface. They describe what they want to build. The platform handles the rest.</p><p>A developer doesn’t have to understand Docker, cloud regions, database provisioning, or compliance requirements. The platform abstracts all of that away. What they do instead is fill out a simple form describing their application, their data classification level, and their expected traffic.</p><p>The platform provisions the infrastructure, configures the deployment pipeline, sets up monitoring, applies security controls, and enforces compliance constraints. All automatically. This is called a golden path. It’s the organization-approved way of building and deploying applications. Following it is not just recommended. It’s the easiest path.</p><h3>How Much Faster Can Your Organization Actually Go?</h3><p>Developers working within a good platform move from code to production in hours. In organizations still managing fragmented DevOps, it takes days or weeks.</p><p>That difference compounds. A developer using the platform can build a feature, deploy it, measure how it performs in production, and iterate. All in a day. The same developer in a fragmented environment spends half their day waiting for infrastructure changes, coordinating with operations teams, and working around inconsistent systems.</p><p>Over a year, that’s not a 20% productivity difference. Studies show it’s closer to 30%. But the real gain isn’t in speed alone.</p><h3>Can You Really Have Both Speed and Safety?</h3><p>Here’s where most organizations get it wrong. They assume speed and safety are trade-offs.</p><p>Faster deployment means more risk. Shorter release cycles mean less time for review. More independence for developers means less control. Companies built on good platforms disprove all of that.</p><p>The platform enforces governance automatically. When a developer deploys something, the platform checks whether the configuration meets security requirements. It validates that the data handling matches the classification level. It ensures monitoring is enabled before a service goes live. It prevents deployments during peak traffic windows.</p><p>None of this requires a manual checklist. None of it slows the developer down. It just happens. The developer gets speed. The organization gets control. These aren’t conflicting goals anymore.</p><h3>What Does This Look Like in Practice?</h3><p>Compare two companies building the same feature.</p><p>Company A has platform engineering. A developer writes the code, commits it, and the platform automatically runs tests, security scans, and builds the deployment package. The developer clicks deploy. The platform provisions the infrastructure, configures monitoring, and validates that everything meets requirements. The service is live in production in four hours. It’s running with the organization’s standard security configuration, logging policy, and cost controls.</p><p>Company B doesn’t have platform engineering. The same developer finishes the code and opens a request with the operations team. Operations team is working on three other requests ahead of theirs. Three days later, they provision the infrastructure. The developer realizes the database configuration doesn’t match what they expected. Another delay while that gets sorted. The service finally goes live ten days later. It’s running in a configuration that’s slightly different from what other services use. Monitoring is configured manually by the developer. If a problem occurs, it’s not routed to the right team.</p><p>That’s not an edge case. That’s typical.</p><p>Scale this across dozens of features a year, hundreds of developers, and the difference in innovation capacity becomes massive. Company A ships features faster. They experiment more. They recover from failures more quickly. They have better insight into what’s happening in production.</p><p>Company B is constantly explaining why everything takes longer.</p><h3>Why Should You Care About This Now?</h3><p>This shift matters now because enterprises are finally mature enough to do it right.</p><p>The first generation of platform <a href="https://www.tothenew.com/services/digital-engineering?SSESS-randomstr=ran-0.22503012721426786">engineering</a> was expensive and complex. Organizations needed dedicated teams with deep infrastructure expertise. You couldn’t build a platform unless you had the talent and the budget.</p><p>Modern platforms are simpler. The abstractions are cleaner. The tools are better. Building a good internal developer platform is still work, but it’s no longer a massive undertaking. At the same time, the cost of not having a platform is becoming clearer. As organizations scale engineering teams and accelerate product development, fragmented DevOps becomes an escalating tax on productivity.</p><p>Enterprises that invested in platform engineering two years ago are already shipping twice as much as competitors still managing DevOps fragmentation.</p><h3>What Should Your Organization Do?</h3><p>The question for technology leaders isn’t whether to build an internal developer platform. It’s how quickly they can build one.</p><p>The organizations that solve this problem first won’t be faster at a few things. They’ll be systematically faster at everything. They’ll deploy more frequently. They’ll recover from incidents more quickly. They’ll onboard developers faster. They’ll make more confident decisions because they have better data.</p><p>Speed doesn’t just matter for speed’s sake. It’s how organizations discover what customers actually need, respond to changes in the market, and stay ahead of competitors who are still waiting for infrastructure changes to be approved.</p><p>The separation between fast companies and slow ones used to be about whether they had good technology. Now it’s simpler: it’s about whether they organized their engineering systems to make speed the natural consequence of building things. Companies with platform engineering do that. Everyone else is competing at a disadvantage.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=01de6971229d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The $100 Million GenAI Mistake Media Companies Are Making Right Now]]></title>
            <link>https://medium.com/@marketing_87790/the-100-million-genai-mistake-media-companies-are-making-right-now-feea241eef6e?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/feea241eef6e</guid>
            <category><![CDATA[genai]]></category>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Fri, 13 Nov 2026 07:36:07 GMT</pubDate>
            <atom:updated>2026-11-13T07:36:07.256Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Sci fi artificial intelligence tech concept background design" src="https://cdn-images-1.medium.com/max/1024/1*G2KTfai_yh6GHIL23jlWVQ.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p><strong>30% of media companies are now allocating budgets for GenAI implementations.</strong></p><p>They’re training models. Hiring AI teams. Building pipelines. Measuring output. And most of them are still making movies the same way. Just faster.</p><p>The companies that will actually win with <a href="https://www.tothenew.com/services/generative-ai-services">GenAI</a> in the media are thinking about something completely different.</p><h3>The Trap Most Media Companies Fall Into</h3><p>When a new technology arrives in media, the instinct is automatic: how do we use this to do what we already do, but cheaper or faster?</p><p>With streaming, this meant: let’s move traditional TV content to a digital platform. Eventually, it meant: let’s use the platform to understand audiences differently. The second group won.</p><p>With GenAI, the same pattern is emerging.</p><p>Most media companies see GenAI as a production efficiency tool. Faster pre-visualization. Quicker visual effects. Automated editing. Content generation at scale. These things are real. But they’re not where the value is.</p><p>The companies thinking strategically about GenAI aren’t asking: “How do we produce the same content 30% faster?”</p><p>They’re asking: “What new kinds of content, experiences, and revenue can we create because GenAI makes it economically possible?” These are completely different questions. And they lead to completely different outcomes.</p><h3>What GenAI Actually Changes For Media</h3><p>Faster production is the obvious answer. But speed was never the bottleneck for media companies. Budget and time were never the constraint on creativity.</p><p>The real constraints were economic and logistical.</p><p>Before GenAI, personalizing content for different audiences, different languages, different platforms meant hiring teams and managing complexity. Economically, it didn’t make sense unless you had massive scale.</p><p>GenAI changes that equation.</p><p>You can now generate variations that adapt to individual viewers, regions, and platforms without proportional increases in cost. You can create localized experiences without localized production teams. You can experiment with formats that previously would have been too expensive to justify.</p><p>You can power real-time enhancements like instant highlight reels for sports, automatically generated match summaries, dynamic overlays with player statistics. You can generate multilingual content instantly, making global events genuinely accessible rather than just subtitled.</p><p>This isn’t about production efficiency. This is about enabling new business models that didn’t exist before.</p><h3>Where The Real Money Is</h3><p>Most media companies are still measuring GenAI impact by cost savings: “Our production pipeline runs 25% faster.”</p><p>The companies building real advantage are measuring it by new revenue: “We can now reach markets we previously couldn’t serve economically.”</p><p>There are three areas where this is happening:</p><h4><strong>First, hyper-personalization at scale:</strong></h4><p>Media companies have always known that personalized content performs better. But personalizing at scale meant hiring more producers, creating more versions, managing complexity.</p><p>GenAI makes it economically feasible to generate personalized variations for millions of viewers. Not just different marketing campaigns or recommendation feeds. Actual content variations tailored to individual preferences, demographics, and behaviors.</p><p>This sounds like a nice feature. But it changes the business model. Instead of a hit-driven business (a few blockbusters, many failures), you can move toward a hit-and-long-tail model where most content performs better because most viewers see something closer to what they actually want.</p><h4><strong>Second, AI-native content formats:</strong></h4><p>Beyond making existing content faster, GenAI enables entirely new content types that weren’t economically viable before. Interactive narratives that adapt in real-time to viewer preferences. Micro-dramas generated and customized for mobile viewers. Virtual influencers and AI co-creators that can produce regular, fresh content without traditional production cycles.</p><p>These aren’t incremental improvements. They’re new products. New revenue streams. New audiences.</p><h4><strong>Third, unlocking the value already sitting in archives:</strong></h4><p>Traditional media companies have accumulated decades of content. Most of it is difficult to find, monetize, or repurpose.</p><p>GenAI can generate advanced metadata for legacy content, making it searchable and usable. It can create device-optimized versions for modern platforms. It can extract clips, generate summaries, create new content from old material.</p><p>A film library that was slowly becoming less valuable can become actively generative when you can automatically optimize it for every platform, language, and audience segment.</p><h3>The Organizational Shift Nobody Talks About</h3><p>There’s a common theme in how companies successfully adopt new technologies: the ones that win are the ones that change their business first and then adopt the technology.</p><p>The ones that adopt the technology first and hope the business changes usually don’t win.</p><p>Most media companies are building <a href="https://www.tothenew.com/insights/article/can-ai-build-house-for-you">AI</a> teams alongside their existing production teams. They’re treating GenAI as a tool that fits into the old structure. The companies thinking strategically are reorganizing around GenAI. Not replacing creators with AI. But restructuring who does what.</p><ul><li>Maybe the creative director still owns the creative vision. But instead of managing 10 editors, they manage 2 editors and a GenAI pipeline. The workflow is different. The role is different. The economics are different.</li><li>Maybe instead of having one localization team for one market, you have GenAI handling localization with human review for cultural accuracy. You can suddenly reach 50 markets instead of 5.</li><li>Maybe instead of waiting weeks for focus group data, you have real-time insights about which storylines resonate, which characters generate traction, which formats are likely to succeed. This informs creative decisions before you invest in full-scale production.</li></ul><p>The organizational structure changes because the way work actually gets done changes.</p><h3>The Shift From Capex To Opex</h3><p>Traditional media production is capital intensive. You build studios. You buy equipment. You staff production. These are fixed costs.</p><p>GenAI flips this. Instead of capital investments in infrastructure, you’re paying for subscription-based services and cloud consumption. Instead of staffing fixed production teams, you’re managing variable usage of AI tools. This sounds like a cost reduction. But it’s actually a strategic shift.</p><p>Capex-heavy businesses move slowly. You invest millions in a facility and you’re committed to using it for years. You staff a team and you’re committed to paying them regardless of project volume.</p><p>Opex-heavy models with AI and <a href="https://www.tothenew.com/cloud-devops">cloud services</a> are inherently more agile. You can scale up to experiment with a new format. Scale down when you want to focus on other projects. Shift resources across initiatives without the friction of infrastructure investment and team restructuring. The companies that move fastest will be the ones that embrace this shift and actually reorganize their cost structure and operations around it, not the ones that just adopt AI tools while maintaining a capex-first mindset.</p><h3>The Challenge Most Companies Miss</h3><p>If GenAI is so obviously valuable, why aren’t all media companies winning with it?</p><p>The standard answer is: talent shortage, legacy infrastructure, governance complexity. These are real. But they’re not the main problem.</p><p>The main problem is simpler: most companies still don’t know what they’re trying to build. They know GenAI exists. They know they should invest in it. They don’t know what business outcome they’re actually trying to achieve.</p><p>Some companies are clear: “We want to reach 50 markets with localized content at a fraction of traditional cost.” That clarity drives decisions about which tools to build, which teams to hire, which processes to change. Other companies just know: “AI is important and we need to do something.” That uncertainty means they build capabilities that don’t connect to revenue, hire teams that don’t align with strategy, and measure success by technology metrics instead of business outcomes.</p><p>The companies winning with GenAI in media are the ones that started with a clear business objective and then figured out which GenAI capabilities support it.</p><h3>The Strategic Distinction</h3><p>Here’s the choice every media leader faces:</p><p>Adopt GenAI tools and workflows to make existing production faster, or reorganize the business around what GenAI makes economically possible.</p><p>One approach maintains existing margins. The other fundamentally changes the business model.</p><p>One company invests in GenAI pipelines and reduces production time from 6 months to 4 months. Another company invests in GenAI and suddenly reaches 50 markets where they previously reached 5, or creates micro-content formats at scale that their previous cost structure never allowed.</p><p>The difference isn’t the tools. The difference is whether you’re thinking about GenAI as an optimization of the old model or as an enabler of a new model. The companies that figure out the new model first will own their market. The ones that focus on making the old model 30% faster will be defending against the ones who did.</p><h3>What Actually Matters Right Now</h3><p>Here’s what media leaders should think about:</p><p>Don’t start by asking: “How do we use GenAI in production?” Start by asking: “What new business outcome do we want to enable?”</p><ul><li>Do you want to reach new geographic markets? Personalization at scale enables that. Localization becomes economically viable.</li><li>Do you want to create content formats that appeal to younger, mobile-first audiences? AI-native interactive content becomes possible.</li><li>Do you want to increase the value of your content library? Archive optimization and derivative content generation become possible.</li></ul><p>Once you know the business outcome, then ask: “Which GenAI capabilities support this?”</p><p>Then ask: “How does our organization need to change to actually deliver this?” The technology is powerful. But the organizations that win won’t be the ones with the fanciest AI tools.</p><p>They’ll be the ones with clear business objectives and the organizational discipline to build around them.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=feea241eef6e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Developer’s Job Isn’t Going Away, It’s Just Moving Upstairs]]></title>
            <link>https://medium.com/@marketing_87790/the-developers-job-isn-t-going-away-it-s-just-moving-upstairs-12a96f23c547?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/12a96f23c547</guid>
            <category><![CDATA[developer]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 06:50:19 GMT</pubDate>
            <atom:updated>2026-11-12T06:50:19.741Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Rear view of programmer working all night long" src="https://cdn-images-1.medium.com/max/1024/1*u13fMgHjT-mbXqqvIeHiqg.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p><strong>Your developers have GitHub Copilot, Cursor, Claude.</strong></p><p>Code generation has never been faster. Features that took weeks now take days. Tests write themselves. Complex functions become suggestions.</p><p>And everyone’s worried that developers will become redundant. They’re worried about the wrong thing.</p><p>The real question isn’t whether AI can write code faster. The real question is: if writing code is no longer the bottleneck, what should developers actually spend their time on? And the answer makes developers more valuable, not less.</p><h3>The Bottleneck That Just Disappeared</h3><p>For decades, <a href="https://www.tothenew.com/services/product-engineering">software development</a> had one obvious constraint: translating ideas into working code took time.</p><p>That constraint is gone. Or at least, it’s no longer the thing that slows teams down.</p><p>A developer can describe a feature, generate a first implementation, ask for alternatives, write tests for edge cases, understand a codebase they’ve never seen, and iterate without rewriting everything from scratch. The mechanical work that used to consume most of a developer’s day now happens in minutes.</p><p><a href="https://survey.stackoverflow.co/2025/ai">Stack Overflow’s 2025 Developer Survey</a> shows 84% of developers are using or planning to use AI tools. 51% use them daily. The shift from “should we use AI?” to “how do we use AI well?” has already happened.</p><p>But here’s the part that matters: faster code generation doesn’t make architectural decisions less important. A production system still needs someone who understands why this particular solution makes sense. Someone who can evaluate trade-offs. Someone who knows what could go wrong.</p><p>AI can generate the solution. The developer still has to decide whether it’s the right solution. That distinction is becoming central to what developers actually do.</p><h3>From Writing Code To Deciding What Should Exist</h3><p>This is where the real shift happens.</p><p>When building a prototype took months, teams committed carefully. Build something, get it right, ship it and then next project.</p><p>When building a prototype takes a day because AI handles the implementation, the conversation changes completely.</p><p>Teams can experiment more. They can try different approaches. They can test ideas before making large engineering bets. But they can also build unnecessary features. Create technical debt. Add complexity that nobody needs.</p><p>The scarce resource stops being “someone who can write this code” and becomes “someone who knows whether this code should exist at all.” That’s a massive shift in what developers are actually doing.</p><p>A developer using AI isn’t just writing code faster. They’re deciding which ideas are worth building, which prototypes deserve investment, which optimizations matter, and which complicated solutions can be replaced by simpler ones.</p><p>That’s engineering judgment. That’s architecture thinking. That’s the work that actually determines whether software succeeds.</p><p>And it’s exactly what developers should be doing more of.</p><h3>What Developers Actually Don’t Trust</h3><p>Here’s the honest part: developers should be skeptical of AI output.</p><p>Stack Overflow’s survey found 46% of developers distrust AI output. When asked why, 66% said their biggest frustration was getting solutions that were “almost right, but not quite.”</p><p>This is the critical insight. AI can write code that runs. AI can write code that passes tests. AI can write code that looks good. But code that technically works isn’t the same as code that should be in production.</p><p>A solution might be logically correct and architecturally wrong. It might handle the happy path and ignore edge cases. It might solve today’s problem and create tomorrow’s technical debt. It might be secure in obvious ways and vulnerable in subtle ones.</p><p>This is where the developer’s role becomes essential. The strongest model isn’t “AI writes code, developer approves.” The strongest model is “AI accelerates work, developer evaluates, decides, and remains accountable.”</p><p>The developer provides context, challenges assumptions and reviews the implementation, tests what could break and understands the system well enough to know whether this solution belongs in it.</p><p>As AI becomes more capable at code generation, those skills become more valuable. Not less valuable. More.</p><h3>The Measure That Actually Matters</h3><p>Here’s where most organizations measure wrong. They count commits. Pull requests. Story points. Lines of code. Hours saved.</p><p>These metrics were always imperfect. AI makes their limitations obvious.</p><p>If a developer generates twice as much code, that doesn’t mean twice the value. A developer who simplifies a complex system, prevents a production incident, improves reliability, or validates a product idea creates more value than someone who closes more tickets.</p><p>The metric should move from “how much code did you produce?” to “what problem did you solve?”</p><p>Quality. Reliability. Customer impact. Time to value. Complexity removed. Technical decisions that actually improve the system. These are the measures that matter when the mechanical work of code generation is no longer the bottleneck.</p><h3>The Role That’s Emerging</h3><p>There’s a misconception that AI-assisted development makes developers less technical.</p><p>The opposite is possible.</p><p>When machines handle more mechanical work, developers have more capacity for system-level thinking. Understanding architecture. Questioning requirements. Evaluating trade-offs. Designing for failure. Thinking about security. Understanding users.</p><p>These aren’t skills that become irrelevant. They become critical. Because good code that doesn’t belong in a system is worse than no code at all. Fast shipping of the wrong thing is just efficient failure.</p><p>The developer is moving from being primarily a producer of code to being an owner of technical outcomes.</p><p>That’s not a reduction in technical skill. It’s an expansion of what technical skill means.</p><h3>How Organizations Actually Misuse AI</h3><p>Here’s the trap organizations fall into: they give developers AI tools and expect faster ticket closure. That’s backwards.</p><p>Effective AI-assisted development requires strong foundations. Reliable tests. Clear architectural boundaries. Security controls. Code review processes that actually evaluate decisions, not just syntax. Observability that shows whether code is working in production.</p><p>Most importantly, it requires a culture where developers question AI output rather than accepting it because it’s faster.</p><p>Without these foundations, AI just makes it easier to write bad code faster.</p><p>With them, AI becomes leverage. Developers can explore more. Prototype more. Build more thoughtfully. Make better decisions because they have more capacity to consider alternatives.</p><p>The difference between organizations that get real value from AI and organizations that just get faster failure is whether they’ve built the discipline to use it properly.</p><h3>The Distinction That Matters</h3><p>Here’s the choice organizations face:</p><p>Treat <a href="https://www.tothenew.com/insights/article/ai-frontend-development-collaboration-speed-and-culture">AI development</a> tools as a way to produce more code with fewer people, or treat them as a way to free developers to do higher-value thinking.</p><p>One approach speeds up execution. The other changes what developers can actually accomplish. One organization invests in faster code generation and gets faster code generation. Another invests in the thinking and discipline around AI-assisted development and gets engineering teams that can explore more, learn faster, and build with greater leverage.</p><p>The outcome isn’t measured in commits or story points. It’s measured in better decisions, fewer production issues, more innovation, faster learning.</p><h3>The Future That Actually Excites Me</h3><p>I don’t see AI in software development as a story about making developers less important. I see it as a shift in what developers are able to focus on.</p><p>The developers who will thrive aren’t the ones who are fastest at typing code. They’re the ones who understand when not to build something. Who can evaluate trade-offs? Who think about system behavior, not just feature implementation. Who questions whether the first solution is the right solution. Those skills are exactly what AI makes more valuable, not less.</p><p>Companies investing in AI development tools aren’t investing in faster code production. At their best, they’re investing in engineering teams that can think more deeply about problems. Explore more approaches. Make better decisions about what actually matters.</p><p>The craft of engineering isn’t disappearing. It’s finally getting room to breathe.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=12a96f23c547" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Your Customers Are Jumping Between Apps, Websites And Portals]]></title>
            <link>https://medium.com/@marketing_87790/your-customers-are-jumping-between-apps-websites-and-portals-because-your-systems-arent-connected-a62edf659350?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/a62edf659350</guid>
            <category><![CDATA[digital-experience]]></category>
            <category><![CDATA[customer]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Fri, 06 Nov 2026 07:13:51 GMT</pubDate>
            <atom:updated>2026-11-06T07:15:24.536Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Representation of user experience and interface design" src="https://cdn-images-1.medium.com/max/1024/1*tUJpdWQOU5jCC3hIcUNaVg.jpeg" /><figcaption>Image by Magnific</figcaption></figure><p>Your customer portal has all their transaction history. Your support system has their service requests. Your website has their preferences. Your email system has their engagement data.</p><p>But your customer has to remember where each piece lives. They tell the same story to multiple channels. Support can’t see what happened on your website. Your website doesn’t know about their recent purchase.</p><p>This is the hidden tax of digital fragmentation. Every disconnected system adds friction. Every silo multiplies the effort required to do simple things. And fixing it isn’t about choosing better software.</p><h3>The Platform Trap</h3><p>Most enterprises approach digital transformation as a platform problem.</p><p>They think: “We need a better CMS. Or a better commerce system. Or a better analytics tool. Then we’ll have a great customer experience.” This logic is backwards.</p><p>A beautiful platform doesn’t create a good experience if the platform is isolated from everything else. A sophisticated analytics system doesn’t improve conversion if the insights never reach the team that owns the customer journey. A personalization engine that doesn’t have access to unified customer data is just making random guesses.</p><p>The real problem isn’t the platform. The real problem is integration, data, and how information flows through the organization to actually affect what customers see and do.</p><p>An enterprise with three mediocre systems connected together will create a better customer experience than an enterprise with one sophisticated system that sits alone.</p><h3>What Digital Experience Actually Means</h3><p><a href="https://www.tothenew.com/cx">Digital experience services</a> aren’t about design aesthetics or platform features.</p><p>Digital experience is about removing friction from how customers interact with your organization across every touchpoint. It’s about having one consistent view of a customer instead of fragmented views across disconnected systems. It’s about embedding customer insights into the workflows where decisions actually get made.</p><p>This requires three things working together simultaneously:</p><h4><strong>First, you need to understand the customer journey:</strong></h4><p>Not in theory, in reality. Where do customers struggle? Where do they abandon? Where do they spend surprising amounts of time? What information do they need but can’t find?</p><h4><strong>Second, you need connected systems:</strong></h4><p>Content lives in one place but can be used everywhere. Customer data is unified so that every channel understands the same customer. Commerce, support, marketing, and service can all access what they need without recreating information or forcing customers to re-enter details.</p><h4><strong>Third, you need the technical discipline to connect insight to action:</strong></h4><p>The analytics platform identifies what’s working. The team that owns the experience receives that information. They make decisions. The system reflects those decisions immediately.</p><p>Most enterprises skip one or two of these. That’s why their digital investments underperform.</p><h3>The Four Things That Actually Matter</h3><p>Forget about platform capabilities for a moment. Ask yourself these questions instead:</p><h4><strong>Where do customers experience the most friction?</strong></h4><p>This is the starting point. Not “what platform should we use?” but “what’s actually making customers frustrated or slowing them down?”</p><p>For a retail company, it might be that customers can’t find products easily. For a financial services company, it might be that customers have to contact support to answer basic questions about their account. For a B2B company, it might be that customers can’t renew contracts without talking to an account manager.</p><p>Identify the friction. Make that the problem you’re solving.</p><h4><strong>What data would eliminate this friction?</strong></h4><p>If customers struggle to find products, you need to connect product information, inventory, personalization, and search. If customers can’t understand their accounts, you need to consolidate account data, transaction history, and support interactions in one view.</p><p>Don’t collect all data. Collect the data that solves this specific problem.</p><h4><strong>How do you get the insight to the people who can act on it?</strong></h4><p>This is where most digital transformations fail silently.</p><p>You might identify that a customer is about to churn based on their engagement patterns. But if the retention team doesn’t see that signal, it’s useless. You might discover that checkout is abandoning because customers are confused about shipping costs. But if that insight doesn’t reach the product team, nothing changes.</p><p>Build the workflow that connects insight to decision-maker. Make sure the decision-maker has the context and tools to act.</p><h4><strong>How do you measure whether it actually worked?</strong></h4><p>Did friction decrease? Did customers complete the task more easily? Did conversion improve? Did support volume drop? Did retention increase?</p><p>Don’t measure whether you implemented the technology. Measure whether the business outcome changed.</p><h3>The Three-Layer Stack That Actually Works</h3><p>When enterprises build digital experiences that matter, they organize around three things:</p><h4><strong>Layer 1: Strategy and design:</strong></h4><p>This is the thinking that comes before technology. What problem are we solving? Who is the customer? What’s the measurable goal? What does success look like?</p><p>This phase often gets rushed. It shouldn’t. Getting this phase right determines whether your technology investment creates value or just speeds up failure.</p><h4><strong>Layer 2: Content and data systems:</strong></h4><p>Content needs to be structured so it can be used in multiple channels. Customer data needs to be unified so different teams work from the same view. These aren’t separate initiatives. They’re the foundation of everything.</p><p>This is where most enterprises discover their technical debt. Systems that made sense five years ago now create friction. Data that lives in isolated databases now needs to be connected.</p><h4><strong>Layer 3: Technology that connects everything</strong></h4><p>The specific technology matters, but only if it serves the strategy and data architecture you’ve already defined.</p><p>Some organizations use traditional platforms. Others use modular systems built on APIs. Some use a mix. The right choice depends on your business problem and operational complexity, not on what’s trendy.</p><p>The critical thing is that technology connects, not that it’s sophisticated.</p><h3>Why Personalization Matters More Than You Think</h3><p>There’s a gap between how enterprises talk about personalization and how personalization actually works.</p><p>Enterprise leaders say: “We need to personalize experiences.”</p><p>What they often mean: “We want to show customers different marketing messages.”</p><p>True personalization is much bigger. It’s about showing customers the information and options that are relevant to them specifically. It’s about prioritizing their account information based on what they need right now. It’s about surfacing the product recommendations that make sense for their needs and history.</p><p><a href="https://www.mckinsey.com/featured-insights/mckinsey-explainers/what-is-personalization">McKinsey research</a> shows that companies excelling at personalization generate 40% more revenue from personalization activities than average performers. But this isn’t about getting the creative tone right. It’s about having the data and systems to understand what each customer needs and the architectural flexibility to deliver it.</p><p>Most enterprises don’t have the data connectivity or system integration to do real personalization. That’s why they default to demographic segmentation and generic messaging.</p><p>Building true personalization requires solving the data and integration problems first.</p><h3>The Real Enterprise Use Cases</h3><p>Here’s what <a href="https://www.tothenew.com/insights/article/enterprise-website-rebranding-digital-experience-transformation">digital experience transformation</a> looks like in practice:</p><h4><strong>Self-service that actually works:</strong></h4><p>Customers can answer their own questions without contacting support. They can complete transactions without talking to an account manager. They can access their information without email requests.</p><p>This works when you integrate customer data with knowledge systems and make that information accessible through search and intuitive navigation. It reduces support costs. It improves customer satisfaction. It gives customers control.</p><h4><strong>Journeys that remember customers:</strong></h4><p>When a customer arrives at your website, it knows who they are. It knows what they’ve purchased. It knows what they’re interested in. It can show them relevant content without forcing them to search or describe what they want.</p><p>This works when you connect identity systems, transaction history, and behavior data with your digital channels. Customers get a consistent, personalized experience. Conversion improves. Customer lifetime value increases.</p><h4><strong>Content at scale:</strong></h4><p>Large enterprises often struggle with content. They have hundreds of products, dozens of markets, multiple brands, constant updates. Publishing becomes slow and fragile.</p><p>This works when you separate content structure from presentation. When you govern content consistently while allowing local customization. When you can publish changes without rebuilding systems. Enterprises move faster. Content stays fresh. Consistency improves.</p><h3>How To Measure If It’s Actually Working</h3><p>Too many digital transformation projects measure success by implementation metrics: “Did we launch the new website?” “Did we migrate all content?” These are irrelevant. What matters is business impact.</p><p>The progression should be:</p><p>Your experience works better (customers can find what they need, complete tasks without friction, get answers without contacting support).</p><ul><li>Then, customers engage more (they visit more often, spend more time, interact with more content and products).</li><li>Then, customers convert more (they purchase, renew contracts, adopt new products, recommend your company).</li><li>Then, customers stay longer (retention improves, lifetime value increases, churn decreases).</li></ul><p>Measure all and ignore any transformation project that can’t show movement in at least the first two.</p><h3>The Strategic Distinction</h3><p>Here’s the choice every enterprise leader faces when planning digital transformation: Treat digital experience as a platform project (select software, implement features, go live), or treat it as a business outcomes project that uses technology.</p><p>One approach produces websites that work like technology. The other produces experiences that solve customer problems. One enterprise invests millions in a sophisticated platform and wonders why customers still struggle. Another enterprise invests in understanding friction, connecting data, and building workflows and sees measurable improvement in conversion, retention, and satisfaction.</p><p>The difference isn’t the platform. It’s the thinking that precedes it. Organizations that get this distinction right build competitive advantage through their customer experience. Organizations that skip it build elaborate systems nobody cares about.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a62edf659350" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Millions In Data Investment Creates Zero Business Value]]></title>
            <link>https://medium.com/@marketing_87790/why-millions-in-data-investment-creates-zero-business-value-13f30c1a417c?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/13f30c1a417c</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[enterprise-ai]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Wed, 04 Nov 2026 06:22:33 GMT</pubDate>
            <atom:updated>2026-11-04T06:22:33.490Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Forecast expert ceo analyzing investment insights and negotiations details" src="https://cdn-images-1.medium.com/max/1024/1*4_MDFARUQ-N3t1BJAPJ-RA.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p><strong>Your company just spent millions on a data platform. Customer data is consolidated from ten different systems. The infrastructure is solid. Everything is technically ready.</strong></p><p>Then someone asks: “So what have we actually changed?” And nobody has a good answer.</p><p>This is the gap that derails most <a href="https://www.tothenew.com/services/generative-ai-services">enterprise AI</a> initiatives. Having data and having the ability to use data to drive business decisions are completely different things.</p><h3>The Hidden Truth About Enterprise Data And AI</h3><p>There’s a seductive story enterprises tell themselves: if we just consolidate all our data, we can use AI to extract value. This story is mostly wrong.</p><p>Having data doesn’t automatically create value. Having access to advanced <a href="https://www.tothenew.com/insights/article/ai-models-and-token-economy-hidden-cost-ai-adoption">AI models</a> doesn’t automatically create value. Having a beautiful data platform doesn’t automatically create value.</p><p>Value comes from connecting data to actual business decisions. And that’s where most enterprises get stuck. Because the real work isn’t technical. It’s organizational.</p><p>You need to identify which business problems matter most. You need to figure out which data actually solves those problems. You need to ensure the insights reach the people who can act on them. You need to measure whether the actions actually worked. Most enterprises skip this thinking. They collect data and hope value emerges. It doesn’t.</p><h3>Why Data Investments Fail Silently</h3><p>Here’s the pattern most enterprises follow:</p><p>They launch a data platform. Teams can now access more information. Engineers celebrate technical achievement. Then… nothing changes.</p><p>Why? Because nobody connected the data to actual decisions.</p><p>A company might consolidate customer data from CRM, transactions, website behavior, and support interactions. This is technically sophisticated. But if nobody uses this consolidated view to make different decisions about customer retention, then it’s just expensive storage.</p><p>Another company builds a predictive analytics capability. The model can identify which customers are likely to churn. But if there’s no workflow that routes this information to the retention team, who then takes action, the prediction is just noise.</p><p>A third company implements AI-powered search for internal knowledge. It works beautifully technically. But if employees still email experts instead of searching, because the search wasn’t embedded in where they actually work, then adoption stalls.</p><p>The data is good. The AI is capable. The infrastructure works. But the value is zero because the insight never reaches the decision-maker, or the decision-maker doesn’t have a process to act on it.</p><h3>The Five Questions That Actually Matter</h3><p>Forward-thinking organizations ask a different set of questions before they invest in data and AI.</p><ul><li>Instead of: “What data do we have?” They ask: “Which business problem should we solve first?”</li><li>Instead of: “Can we build a model for this?” They ask: “Do we have the data needed to support this solution?”</li><li>Instead of: “How sophisticated can our AI be?” They ask: “What needs to change in how we work to actually use this insight?”</li><li>Instead of: “Did we implement the technology?” They ask: “Who owns the business outcome, and how do we measure success?”</li></ul><p>These questions sound simple. They’re not. But answering them correctly changes everything.</p><p>The organizations that ask these questions first don’t waste money building models for problems that don’t matter. They don’t create insights that nobody acts on. They don’t measure success by technology adoption. They measure success by business impact.</p><h3>The Four-Step Path That Actually Works</h3><p>Most organizations approach data and AI as technology projects. They should approach them as business-outcome projects that use technology. This shift changes how you plan and prioritize.</p><h4><strong>Step 1: Start with a business problem, not a technology solution:</strong></h4><p>Don’t ask: “What can we do with AI?” Ask: “What’s making our customers leave?” or “What’s slowing down our operations?” or “Where are we losing money?”</p><p>Be specific. “Improve customer retention” is too vague. “Reduce customer churn for our high-value segment from 8% to 5% annually” is specific.</p><h4><strong>Step 2: Identify which data actually solves this problem:</strong></h4><p>Now that you know the problem, figure out what information matters. For customer churn, this might be purchase history, support interactions, website engagement, and customer feedback. Don’t collect everything. Collect what’s relevant to this specific problem.</p><h4><strong>Step 3: Connect insight to action:</strong></h4><p>Here’s where most organizations fail. They build the insight and forget to build the workflow.</p><p>The insight might be: “This customer is likely to churn.” The workflow is: “Alert the account team, prioritize follow-up, and recommend intervention options.”</p><p>Without the workflow, the insight has zero value. With the workflow, the insight drives action.</p><h4><strong>Step 4: Measure whether it worked:</strong></h4><p>Compare the baseline (before the initiative) with the result (after implementation). Did <a href="https://www.tothenew.com/insights/article/ai-models-and-token-economy-hidden-cost-ai-adoption">churn</a> actually decrease? Did resolution time improve? Did revenue increase?</p><p>Don’t measure whether you implemented the technology. Measure whether the business outcome changed. If it worked, ask: can we apply this approach to other problems? If it didn’t work, ask: why not? What would need to change?</p><h3>The Reusability Question</h3><p>Here’s what separates organizations that build sustainable data and AI capabilities from organizations that build one-off experiments:Ad-hoc organizations: Build an AI model for this specific problem. Measure success. Move on.</p><p>Strategic organizations: Build an AI model for this specific problem. Measure success. Then ask: can we apply this capability to other business use cases? Can we make it reusable? The difference is huge. One organization solves problems one at a time forever. The other organization builds capabilities that solve multiple problems over time.</p><p>This requires thinking about data infrastructure, governance, and processes as platforms, not as one-time projects.</p><h3>The Three Things That Have To Align</h3><p><a href="https://www.tothenew.com/data-services">Data</a> and AI create value when three things align simultaneously:</p><h4><strong>First, you have relevant, reliable data:</strong></h4><p>This doesn’t mean perfect data. It means data good enough for your specific use case, accessible to the right systems and users, and understood well enough to use confidently.</p><h4><strong>Second, you have clear governance and ownership:</strong></h4><p>Someone owns the outcome. Someone owns the data quality. Someone owns the decision of how to use the insight. Without clear ownership, accountability disappears and nothing gets done.</p><h4><strong>Third, the insight is embedded in how work actually gets done:</strong></h4><p>The recommendation reaches the decision-maker. The workflow includes the AI output. The team has been trained to use it. It’s not an afterthought. It’s part of normal operations. When all three align, value happens. When any one is missing, the initiative stalls.</p><h3>Why This Matters Right Now</h3><p>Enterprise AI is moving past the experimentation phase. Organizations are asking harder questions about ROI. They’re tired of pilot projects that don’t scale. They want to know: how do we actually get business value from data and AI investments?</p><p>The answer isn’t more technology. It’s better thinking about which problems matter, which data solves them, and how to connect insights to decisions.</p><p>Organizations that figure this out will extract massive value from their data investments. Organizations that skip this thinking will keep investing in technology that doesn’t move the needle.</p><h3>The Strategic Distinction</h3><p>Here’s the choice every enterprise leader faces:</p><p>Treat data and AI as technology initiatives that happen to involve your business, or treat them as business initiatives that happen to use technology. One approach drives technical adoption metrics. The other drives business outcomes.</p><p>The organizations that get this distinction right are the ones that will actually turn their data investments into competitive advantage. The ones that get it wrong will have nice platforms and frustrated business leaders wondering why nothing changed.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=13f30c1a417c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Gap Nobody Talks About: Perfect Infrastructure, Frustrated Customers]]></title>
            <link>https://medium.com/@marketing_87790/the-gap-nobody-talks-about-perfect-infrastructure-frustrated-customers-918122f91fad?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/918122f91fad</guid>
            <category><![CDATA[digital-experience]]></category>
            <category><![CDATA[customer-experience]]></category>
            <category><![CDATA[infrastructure]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Fri, 30 Oct 2026 08:07:00 GMT</pubDate>
            <atom:updated>2026-10-30T08:07:00.734Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Representation of user experience and interface design" src="https://cdn-images-1.medium.com/max/1024/1*bHU_eus46WnkeFSGp6F0-g.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p><strong>Your infrastructure team reports that everything is working perfectly. Uptime is 99.9%. All APIs are responding. Database performance is solid.</strong></p><p>Meanwhile, your customer service team is getting complaints. Pages are slow. Checkout is frustrating. Mobile experiences feel broken.</p><p>Who’s right? Everyone is. And that’s the core problem.</p><h3>The Gap Between System Health And Customer Experience</h3><p>For decades, enterprises measured digital success by one metric: is the system up? This made sense when systems were monolithic and users had limited options. If your system was up, you were winning. Then the web got complicated. Mobile became essential. Users got intolerant of friction. Competitors are always one click away.</p><p>Now “the system is up” doesn’t mean much. A system can be technically healthy while delivering frustrating experiences.</p><p>Your checkout page might load successfully, but load slowly on mobile networks. Your search might work technically, but take three seconds to return results. Your personalization might be delivering the right content, but creating performance problems for users. The infrastructure team reports success. The business team wonders why conversion is declining.</p><p>This gap is where most <a href="https://www.tothenew.com/">digital transformation</a> gets stuck. Organizations have great visibility into system health. They have almost no visibility into customer experience. Forward-thinking organizations are bridging this gap by measuring what actually matters: how customers experience applications in the real world, not how healthy the infrastructure is.</p><h3>Three Truths About Digital Experience That Change Everything</h3><h4><strong>1. Real users experience conditions your test environment never creates:</strong></h4><p>Your quality assurance team tests the application on a modern laptop with fast internet. Your customers use it on a four-year-old phone on a slow network. They use it at 2am when the server is handling millions of requests. They use it while commuting. Your test environment shows everything working perfectly. Real users face slow interactions, failed transactions, and frustrating experiences.</p><p>This isn’t a technology problem. It’s a visibility problem. You can’t improve what you’re not measuring.</p><h4><strong>2. Performance problems directly impact business outcomes:</strong></h4><p>It’s tempting to treat slow pages as a technical problem for the engineering team to fix someday. But slow pages have immediate business consequences. Research shows that every 0.1-second improvement in mobile site speed correlates with measurable increases in conversion. A slow checkout costs revenue. A slow search costs engagement. A failed mobile transaction costs customers.</p><p>When enterprises start connecting experience problems to business impact, prioritization changes immediately. A problem that seemed minor suddenly becomes urgent when you realize it’s directly affecting revenue.</p><h4><strong>3. Friction often accumulates in unexpected places:</strong></h4><p>Most enterprises track individual metrics: page load time, API latency, JavaScript errors. These are useful. But they miss how friction compounds across a journey. A customer experience might be fine at each step, but the cumulative friction across ten steps creates abandonment. Or a personalization system might be delivering great content but creating performance problems. Or an API integration might be working but taking too long.</p><p>The friction isn’t in any single place. It’s in how everything works together from the customer’s perspective. Organizations that measure customer journeys holistically discover friction nobody noticed when looking at isolated metrics.</p><h3>The Intelligence Question: Why Measure Experience At All</h3><p>Here’s what separates enterprises that optimize digital experiences from enterprises that just hope things work:</p><p>Struggling enterprises measure: Is the system up? Is it fast enough? Are there errors? These are binary questions. Yes or no. The system either works or it doesn’t.</p><p>Forward-thinking enterprises measure: Which customers face which friction? Where does the journey break down? Which problems affect revenue? Which optimizations would have the biggest impact?</p><p>These are intelligence questions. They require data, analysis, and decision-making. The first approach is monitoring. The second approach is business intelligence.</p><p>Most enterprises are doing monitoring. They’re asking the wrong questions and wondering why their improvements don’t move the needle.</p><h3>What Forward-Thinking Organizations Actually Measure</h3><p>Organizations that consistently improve digital experience focus on three things:</p><h4><strong>1. They measure real user behavior, not just system metrics:</strong></h4><p>Instead of just knowing “the page loads in 2 seconds,” they know “the page loads in 2 seconds for users on fast networks but 8 seconds for users on mobile networks, and those users abandon 40% of the time.”</p><p>This specificity matters. It tells you exactly who to optimize for and whether the optimization will actually move the needle.</p><h4><strong>2. They connect experience problems to business outcomes:</strong></h4><p>Instead of just knowing “the checkout page has JavaScript errors,” they know “JavaScript errors on the checkout page affect 8% of transactions and cost us $50,000 per month in lost revenue.”</p><p>Now the problem has business priority. Now engineering knows why this matters. Now stakeholders understand the investment case for fixing it.</p><h4><strong>3. They measure full journeys, not isolated metrics:</strong></h4><p>Instead of just measuring individual page load times, they measure whether customers can complete their goals. Can they find what they’re looking for? Can they complete checkout? Can they access their account?</p><p>Journey success is what matters to business. Individual metrics only matter if they affect journey success.</p><h3>The Hidden Friction Organizations Miss</h3><p>Here’s where most organizations struggle: they optimize for average performance. They aim for pages to load in under 3 seconds. They achieve it. Problem solved.</p><p>Except half their users are experiencing pages that load in 5 seconds. They just don’t know it because everyone sees the average.</p><p>Forward-thinking organizations don’t optimize for average. They optimize for worst-case. They ask: which users face the most friction? Which geographic regions? Which devices? Which networks?</p><p>Then they prioritize problems by impact. A problem affecting 5% of users on slow networks might have a bigger business impact than a problem affecting 50% of users on fast networks if those 5% happen to be high-value customers on mobile devices. This requires visibility into user segments, not just averages.</p><h3>The Experimentation Insight</h3><p>Here’s something that separates mature organizations from developing ones: Developing organizations: Make a change, measure if it went faster, deploy it.</p><p>Mature organizations: Make a change, measure if it went faster AND if customers like it AND if conversion improved AND if retention improved. They run experiments. They measure multiple dimensions. They understand that sometimes making something faster makes it worse if it breaks functionality or changes behavior users don’t like.</p><p>This requires measuring experience, not just performance.</p><h3>Where To Start</h3><p>Organizations adding digital experience visibility typically start with:</p><ul><li><strong>Identify your critical journeys: </strong>which journeys affect revenue or customer satisfaction the most? Start there. Don’t try to measure everything. Measure what matters.</li><li><strong>Establish baselines: </strong>measure how your critical journeys perform right now, across different devices, networks, locations, and customer segments. Create benchmarks.</li><li><strong>Find the friction:</strong> analyze where customers struggle. Which steps take too long? Which have high abandonment? Which have errors? Prioritize problems by business impact.</li><li><strong>Experiment and measure: </strong>when you make changes, measure whether they actually improved the journey from the customer’s perspective. Don’t measure just speed. Measure success, satisfaction, and business outcomes.</li><li><strong>Repeat continuously:</strong> this isn’t a one-time project. As you improve, new friction appears. As your application changes, performance characteristics change. Continuously measure and optimize.</li></ul><h3>Why This Matters Right Now</h3><p><a href="https://www.tothenew.com/cx">Digital experience</a> is becoming the primary competitive differentiator.</p><p>If your checkout is slower than competitors’, you lose conversions. If your mobile experience is frustrating, you lose engaged users. If your search is slow, you lose engagement. Competitors that can see experienced problems and fix them quickly will outpace competitors that are guessing.</p><p>Organizations that measure experience and continuously optimize will deliver better outcomes than organizations that measure system health and hope it correlates with customer satisfaction. The gap between “our systems are healthy” and “our customers are satisfied” is where competitive advantage lives right now.</p><h3>The Strategic Choice</h3><p>Here’s the choice every digital leader faces:</p><p>Continue measuring system health and hoping it correlates with customer experience, or start measuring customer experience and using those insights to drive business outcomes. One approach is cheaper and simpler initially. The other generates better results and compounds over time.</p><p>The organizations that will win in 2027 are the ones that started measuring and <a href="https://www.tothenew.com/insights/case-studies/scripta-insights">optimizing customer experience</a> in 2026.</p><p>Not because they have perfect technology. Because they understand what matters: not whether systems are working, but whether customers can accomplish their goals and have experiences that keep them engaged.</p><p>That understanding, combined with continuous measurement and optimization, is where digital transformation actually happens.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=918122f91fad" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Wrong Way To Choose An AI Model (And What Actually Works)]]></title>
            <link>https://medium.com/@marketing_87790/the-wrong-way-to-choose-an-ai-model-and-what-actually-works-9ef8f46d563d?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/9ef8f46d563d</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-model]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Thu, 29 Oct 2026 05:01:43 GMT</pubDate>
            <atom:updated>2026-10-29T05:01:43.697Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Artificial intelligence and neural network visualization" src="https://cdn-images-1.medium.com/max/1024/1*v1Pk7hzDdDPWuaHu-KJB7A.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p><strong>You’re comparing AI models based on benchmark scores. Everyone’s doing it. Nobody’s happy with the results.</strong></p><p>The model that wins on the leaderboard struggles with your proprietary data. The cheap model needs so many retries that it ends up costing more than the premium one. The high-performance model takes too long for your use case. You’re paying for capability you don’t need.</p><p>This is the wrong problem to solve.</p><h3>The Shift From “Best Model” To “Right Model”</h3><p>For the last couple years, the model selection question was simple: “Which model scores highest on public benchmarks?”</p><p>That was a reasonable question when there were only a few capable models and the differences between them were massive.</p><p>In 2026, that question is almost useless. Why? Because there are now dozens of capable models. They’re specialized. They’re fast. They’re cheap. The differences aren’t about which one is objectively “best.” They’re about which one is best for your specific work.</p><p>This changes everything. The old model selection process looked like: Pick the highest-performing model on a leaderboard, use it everywhere, pay per token, and hope it works.</p><p>The new process looks like: Understand your actual work, match models to specific tasks, measure true cost including retries and failures, and build flexibility to change as new models emerge.</p><p>This shift from “best model” thinking to “right model for this task” thinking is what separates enterprises that get AI economics right from enterprises that keep getting surprised by costs.</p><h3>What Gets Measured Gets Done: Cost Per Outcome vs. Cost Per Token</h3><p>Here’s where most enterprises go wrong: they <a href="https://www.tothenew.com/insights/article/ai-models-and-token-economy-hidden-cost-ai-adoption">measure AI cost in tokens instead of outcomes</a>.</p><p>A token is an abstract unit. An outcome is something that matters to the business.</p><p>Example: You’re extracting information from documents using an AI model. You measure cost per token. Model A costs $0.0005 per 1,000 tokens. Model B costs $0.003 per 1,000 tokens. Model A seems six times cheaper, so you choose it.</p><p>Then you deploy it and discover Model A makes more mistakes. It misses important information 15% of the time. Now you need human review for failures. Some extractions need to be redone. The cheap model has become expensive because you’re measuring the wrong thing.</p><p>The real cost is per successful extraction. If Model A costs half as much per token but requires twice as much human review, it’s actually more expensive.</p><p>Smart enterprises shift to measuring cost per outcome: “What does it cost to complete this task successfully, including retries, validation, and human intervention?” This usually reveals something counterintuitive: <a href="https://www.tothenew.com/insights/article/cheap-vs-superior-intelligence-ai-model-selection">the cheaper model is often more expensive</a>. Or a faster model saves money because it completes more tasks in parallel. Or a model specialized for your domain costs more per token but succeeds on the first try.</p><p>The costs align when you measure what actually matters.</p><h3>The Benchmark Shortlist vs. The Production Reality</h3><p>Public benchmarks are useful as a starting point. They narrow the field from thousands of models to dozens.</p><p>But benchmarks don’t tell you whether a model works for your actual work.</p><p>Here’s why: benchmarks test general capability. They test whether a model can reason, follow instructions, and produce output. They don’t test whether a model understands your terminology, handles your document formats, works with your APIs, or succeeds on the specific edge cases you encounter. A model can dominate public benchmarks yet struggle with your work. And a model that’s weaker on benchmarks can outperform in production because it handles your specific scenarios better.</p><p>The enterprises getting this right do something critical: they test models on their actual work before deciding.</p><p>This means:</p><p>Taking real examples from production. Not sanitized versions. Real customer data, real edge cases, real complications.</p><p>Testing multiple models under identical conditions. Same prompts. Same context. Same tools. Same evaluation criteria. If you change the test between models, you’re measuring the test, not the models.</p><p>Measuring what matters for the task. For translation, accuracy matters. For customer summarization, speed matters. For fraud detection, reliability matters. Different weights for different work. Running tests including failure cases. If the model struggles on a particular type of document or request, your production will too.</p><p>The pattern: benchmark scores get you a shortlist. Production testing determines the winner.</p><h3>The Multi-Model Architecture Insight</h3><p>Here’s the insight that changes everything: you don’t need one model that’s great at everything. You need different models optimized for different tasks.</p><ul><li>A routine classification task doesn’t need a frontier model. Use a faster, cheaper model.</li><li>A creative writing task needs more reasoning capability. Use a stronger model.</li><li>A real-time chat needs minimal latency. Use a fast model even if it’s slightly weaker.</li><li>A complex document analysis needs accuracy. Use a stronger model even if it’s slower or more expensive.</li></ul><p>This seems obvious when stated simply. But most enterprises still use one model everywhere because it’s simpler operationally.</p><p>Forward-thinking organizations are building slightly more complex architectures that let them route work to the right model for each task. Routine work goes to efficient models. Complex work goes to stronger models. This improves both economics and outcomes.</p><p>Is it more complex operationally? Yes. Do the economics justify the complexity? Usually yes, especially as the volume of work grows.</p><h3>What Actually Matters in Model Selection</h3><p>When evaluating which model to use for a specific task, focus on what moves the needle:</p><ul><li><strong>Accuracy</strong>: Does it produce correct results? Most important for financial analysis, medical decisions, contract review. Less important for brainstorming.</li><li><strong>Latency</strong>: How fast does it respond? Critical for real-time interactions. Less important for batch processing.</li><li><strong>Cost</strong>: What does a completed task cost, including retries and validation? This is usually the differentiator.</li><li><strong>Reliability</strong>: How often does it fail or need escalation? High-stakes work needs high reliability.</li><li><strong>Following instructions</strong>: Can it use the tools and data you give it? Can it produce structured output? Can it work with your APIs?</li><li><strong>Context handling</strong>: Does it use provided information effectively? Some models get confused by lots of context. Others thrive with it.</li><li><strong>Security</strong>: Where is your data processed? This matters for regulated industries or proprietary information.</li><li><strong>Flexibility</strong>: Can you move to another model if needed? This protects against vendor lock-in and lets you optimize as new models emerge.</li></ul><p>Different tasks weigh these differently. A voice assistant prioritizes latency. Contract analysis prioritizes accuracy and traceability. High-volume classification prioritizes cost and throughput.</p><p>There’s no universal winner. There’s only the best fit for defined requirements.</p><h3>The Five-Step Process That Actually Works</h3><p>Enterprises that consistently pick the right models follow a clear process:</p><h4><strong>Step 1: Start with work, not models:</strong></h4><p>Map out what you’re trying to do. Extracting information? Classifying content? Generating creative output? Analyzing documents? Each has different requirements. Don’t start by comparing models. Start by understanding work.</p><h4><strong>Step 2: Set hard thresholds:</strong></h4><p>Define minimum requirements. The model must be at least 95% accurate. Must respond in under two seconds. Must cost under $0.01 per task. Must be HIPAA-compliant. These aren’t preferences. These are requirements. A model that fails a requirement shouldn’t win because it scores higher elsewhere.</p><h4><strong>Step 3: Build a real evaluation set:</strong></h4><p>Use actual examples from your work, including your hardest cases. Include edge cases that have caused problems before. Test multiple models on the same examples under identical conditions. This is usually where benchmark leaders and production winners diverge.</p><h4><strong>Step 4: Measure true cost per outcome:</strong></h4><p>Don’t just measure token cost. Measure what it costs to complete the task successfully. Does the model need retries? Does it need human validation? Does it need escalation for hard cases? Include those in the cost.</p><h4><strong>Step 5: Plan to re-evaluate:</strong></h4><p>Model selection isn’t a one-time decision. New models emerge. Pricing changes. Your understanding of what works improves. Set a schedule to re-evaluate. Some organizations do this quarterly. Some annually. The rhythm depends on how fast your environment is changing.</p><h3>The Architecture Question: Should You Build For Optionality?</h3><p>Here’s a strategic decision: should your system be locked to one model, or flexible to switch models? Locked to one model is simpler operationally. You deploy, you’re done, it works.</p><p>Flexibility to switch models is slightly more complex but gives you options. If a better model emerges, you can use it. If pricing changes dramatically, you can optimize. If a model degrades or becomes unavailable, you have a backup.</p><p>For high-stakes, high-volume, or long-term deployments, optionality usually wins. The operational complexity is worth the flexibility. For simple, low-stakes, or short-term deployments, lock-in usually doesn’t matter.</p><p>The strategic question: how much will this work matter in two years? If a lot, <a href="https://www.tothenew.com/insights/article/choosing-right-cloud-operating-model-scale-agility-and-governance">build for optionality</a>. If it might disappear or change completely, don’t over-engineer.</p><h3>Why This Matters Right Now</h3><p>The AI landscape is moving fast. New models are emerging constantly. Pricing is changing. Capabilities are shifting. Specialized models are proliferating.</p><p>Enterprises that pick a model and lock in are betting that model stays optimal. That’s increasingly risky. Enterprises that pick a model but build for flexibility are betting they can adapt as things change. That’s increasingly valuable.</p><p>This is why model strategy (how do we think about choosing and changing models?) matters more than model selection (which model do we pick?).</p><h3>The Real Competitive Advantage</h3><p>Here’s what separates AI-savvy enterprises from AI-struggling enterprises:</p><p>AI-struggling enterprises choose a model based on leaderboards, deploy it everywhere, measure cost in tokens, and discover later that economics don’t work. AI-savvy enterprises understand their work, test models on real examples, measure true cost per outcome, and build flexibility to optimize as things change.</p><p>One approach is cheaper to implement initially. The other is cheaper to operate long-term. The organizations that get AI economics right in 2026 are the ones doing the work upfront to understand what they’re actually trying to solve, then matching solutions to that work instead of forcing work to match available solutions.</p><p>That’s the shift from “which model should we use?” to “how do we systematically choose the right model for each type of work we do?”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9ef8f46d563d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Media Leaders Winning With AI Aren’t Replacing Creatives, They’re Multiplying Them]]></title>
            <link>https://medium.com/@marketing_87790/the-media-leaders-winning-with-ai-arent-replacing-creatives-they-re-multiplying-them-b72e8e6dd5f1?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/b72e8e6dd5f1</guid>
            <category><![CDATA[media]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Thu, 22 Oct 2026 05:48:06 GMT</pubDate>
            <atom:updated>2026-10-22T05:48:06.066Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Abstract ai background glowing globe neon lines" src="https://cdn-images-1.medium.com/max/1024/1*NL-9VstdoJwZezAGwZnCpg.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p><strong>Your streaming platform just launched 200 new titles this quarter. Your newsroom published 500 stories. Your gaming studio shipped updates for 12 games simultaneously.</strong></p><p>None of that was possible five years ago. And none of it was possible without <a href="https://www.tothenew.com/insights/article/what-is-ai">AI</a>. But here’s what’s interesting: the media organizations doing this aren’t the ones replacing creative teams. They’re the ones multiplying what creative teams can do.</p><p>This distinction matters. Because the future of media isn’t about choosing between human creativity and AI. It’s about combining them so effectively that both become more valuable.</p><h3>The Fundamental Shift In Media</h3><p>For most of the media’s history, the bottleneck was production capacity. You had a fixed number of writers, editors, producers, directors. That number determined how much content you could make.</p><p>AI changes this equation. Not by replacing creatives. By changing what the constraints are.</p><p>A writer can now spend less time on research and more time on the story. An editor can spend less time on manual tasks and more time on judgment. A producer can coordinate across teams without managing logistics. A localization team can expand to dozens of languages without proportionally expanding headcount.</p><p>The bottleneck moves from “how many creatives do we have” to “how do we help creatives work at their best.” The media organizations winning with AI right now understand this shift. They’re not automating creatives out. They’re automating the work around creativity so creatives can focus on what they do best.</p><h3>What Forward-Thinking Media Leaders Are Actually Doing</h3><p>The streaming platforms, broadcasters, and publishers scaling AI fastest share a consistent approach: They start small and focus on specific problems.</p><p>Instead of “let’s <a href="https://www.tothenew.com/services/generative-ai-services">implement AI across the organization</a>,” they ask: “What takes our team the most time? What work is repetitive? What would free up time for more strategic work?”</p><p>They identify operational bottlenecks first. Metadata tagging, caption generation, transcript creation, content summarization. These aren’t creative work. They’re the work around creativity.</p><p>Then they automate these tasks. This immediately frees up time and budget. The efficiency gains fund the next step and they reinvest into creativity.</p><p>Once they’ve automated the operational overhead, they use that freed-up capacity differently. Instead of hiring more people to do the same work faster, they invest in deeper creative work. Longer format content. More experimentation. Higher quality production. They layer in audience intelligence.</p><p>As they optimize operations, they build the ability to understand their audience better. Not in abstract demographic terms, but in behavioral terms. What does this specific viewer engage with? What keeps them watching? What makes them come back?</p><p>This intelligence becomes a collaborative tool for creators. It’s not prescriptive (“make this kind of content”). It’s descriptive (“here’s what resonates with your audience”).</p><h3>Three Ways AI Actually Multiplies Creative Power</h3><h4><strong>First, AI handles the research and ideation acceleration.</strong></h4><p>A writer used to spend days researching a story. Now AI can analyze trends, pull relevant information, identify gaps, and present options in hours. The writer still writes the story. But they start from a much stronger foundation.</p><p>A content producer used to spend weeks ideating new show concepts. Now AI can analyze what’s working, identify audience demand, spot opportunities, and suggest directions. The producer still decides. But they decide from a deeper pool of possibilities. The creative judgment remains human. The speed and foundation improve.</p><h4><strong>Second, AI handles the production overhead so creators focus on creation.</strong></h4><p>Video editors used to spend days on manual tasks: organizing footage, color correction, audio balancing, captioning, metadata tagging. Now AI handles much of this. The editor focuses on the editorial choices that matter: pacing, storytelling, emotional impact.</p><p>Transcription used to require hours of manual work. Now AI generates transcripts instantly. Humans still review for accuracy and tone. But they’re editing, not transcribing. Writers can focus on narrative instead of formatting. Producers can focus on vision instead of logistics.</p><h4><strong>Third, AI handles personalization at scale.</strong></h4><p>Traditional media worked on broadcast logic: make one version of something and send it to everyone. The internet enabled targeting: make versions for segments and send the right version to the right segment.</p><p>AI enables something different: personalization. Millions of viewers, millions of variations. Recommendation systems that understand what this specific viewer wants to watch. Dynamic thumbnails that change based on predicted engagement. Notifications timed to when this person is most likely to engage.</p><p>This isn’t replacing editorial judgment. It’s making editorial judgment scalable. A single editor’s insight can now influence the experience of millions of viewers without the editor having to manually create millions of variations.</p><h3>The Localization Opportunity Nobody’s Maximizing Yet</h3><p>One of the highest-impact uses of <a href="https://www.tothenew.com/insights/article/article-ai-media-entertainment-use-cases">AI in media</a> is something most organizations haven’t figured out at scale yet: localization.</p><p>Historically, going global meant massive costs. Dubbing a show into 10 languages required hiring voice actors in 10 countries, coordinating production schedules, managing quality. It was expensive and slow.</p><p>AI changes this. Sophisticated voice synthesis and dubbing technology can now adapt content to different languages while preserving the original performance. Not perfectly, and not for everything. But well enough that it’s viable. This opens a competitive advantage: the ability to reach global audiences on a timeline that used to be impossible.</p><p>The media organizations that figure out how to scale localization effectively will reach markets faster than competitors who are still doing traditional dubbing.</p><h3>Why Technology Foundation Matters (But Isn’t Obvious)</h3><p>Here’s where most media organizations struggle: they invest in AI tools but don’t see proportional returns. The problem is rarely the AI. It’s usually the foundation underneath.</p><p>If your data is scattered across different systems, AI can’t see the complete picture. If your infrastructure doesn’t support real-time processing, personalization becomes reactive instead of proactive. If you don’t have the engineering capacity to integrate AI into existing workflows, AI stays siloed.</p><p>Smart organizations align three things:</p><ul><li>Their AI investments with their data strategy. If you can’t access unified audience data, personalization AI becomes guesswork.</li><li>Their AI tools with their production infrastructure. If AI recommendations can’t flow into your actual distribution system, they don’t matter.</li><li>Their AI capabilities with their team’s skills. If nobody on the team knows how to work with AI outputs, the best tools in the world won’t help.</li></ul><p>The organizations that do this well don’t necessarily have more advanced AI. They have better alignment between AI, data, and operations.</p><h3>The Revenue Opportunity, Not Just Efficiency</h3><p>Most discussions about AI in the media focus on cost savings. Faster production. Fewer people for the same output. This misses the bigger opportunity. AI doesn’t just reduce costs. It enables growth.</p><p>Better personalization means higher engagement. Higher engagement means more watch time, more content consumption, more customer lifetime value.</p><p>Better audience intelligence means smarter investment decisions. You know which types of content drive retention. Which drive new subscriptions. Which drive advertising value. This changes how you allocate budget.</p><p>Faster localization means global reach on timelines competitors can’t match. This is a market expansion opportunity, not just an efficiency improvement. The organizations that treat AI as a growth tool (not just an efficiency tool) generate more value than organizations that treat it as cost reduction.</p><h3>The Culture Shift That Actually Matters</h3><p>Here’s what separates media organizations that successfully integrate AI from ones that don’t: Successful organizations treat AI as a tool for creatives, not a replacement for them.</p><ul><li>They invest in training. Editors learn how to work with AI-assisted editing tools. Writers learn how to use AI research capabilities. Producers learn how to interpret AI-generated audience insights.</li><li>They keep humans in the loop for judgment calls. AI can suggest. Humans decide. Editorial integrity remains human-driven.</li><li>They celebrate the time saved and reinvest it. When AI automates metadata tagging, they don’t lay off the person who did tagging. They retrain them to do something more strategic.</li><li>They measure success by creative output and audience impact, not by automation percentage. The question isn’t “how many manual tasks did we automate?” It’s “did we make better content faster and reach more people?”</li></ul><h3>What The Next 18 Months Will Look Like</h3><p>By late 2026 and into 2027, the competitive advantage in media will come from how <a href="https://www.tothenew.com/brochure/to-the-new-gen-ai-for-business.pdf">effectively organizations integrate AI into their operations</a>.</p><p>The organizations that moved fast will have built operational muscle. They’ll be producing more content, reaching more global markets, personalizing experiences at scale, and making smarter investment decisions. The organizations that waited will still be figuring out whether to start.</p><p>This isn’t a “first mover” advantage that locks competitors out forever. But it is an 18-month window where the organizations moving now build capabilities that become harder to replicate later.</p><h3>The Strategic Choice</h3><p>For media leaders, the choice isn’t whether to use AI. Most competitors will figure that out eventually. The choice is whether to use AI as a tool to multiply human creativity or as a replacement for it.</p><p>The organizations choosing multiplication will build deeper creative capacity. More output, better quality, faster time to market and greater reach. The organizations choosing replacement will squeeze costs out of production until they hit a ceiling.</p><p>One path compounds creativity. The other exhausts it.</p><p>The media leaders winning right now are on the compounding path. They’re using AI to do more with their existing creative talent, then investing the efficiency gains back into even stronger creative work. That’s how you build a sustainable AI-first media enterprise.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b72e8e6dd5f1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When AI Takes Action: The Accountability Question Your Board Will Ask]]></title>
            <link>https://medium.com/@marketing_87790/when-ai-takes-action-the-accountability-question-your-board-will-ask-07450876a211?source=rss-0b5fddc2d7a5------2</link>
            <guid isPermaLink="false">https://medium.com/p/07450876a211</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ai-takes-action]]></category>
            <category><![CDATA[ai-agent]]></category>
            <dc:creator><![CDATA[TO THE NEW Insights]]></dc:creator>
            <pubDate>Mon, 19 Oct 2026 04:59:15 GMT</pubDate>
            <atom:updated>2026-10-19T04:59:15.354Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="3d rendering of biorobots concept" src="https://cdn-images-1.medium.com/max/1024/1*h4n6Ps2GlzsUDdqwENdE-A.jpeg" /><figcaption>Image by <a href="https://www.magnific.com/app">Magnific</a></figcaption></figure><p><strong>Your AI agent just made a decision that changed a customer’s account.</strong></p><p>It wasn’t a recommendation. It wasn’t a summary. It was an action. Money moved. Data changed. A transaction happened. Now someone asks: “Who authorized that?”</p><p>This is the question every enterprise will face in 2026 as <a href="https://www.tothenew.com/insights/article/what-is-ai">AI</a> moves from answering questions to taking actions. And it’s not a technical question. It’s a governance question. It’s a legal question. It’s a board-level question.</p><h3>The Accountability Gap That Haunts Agents</h3><p>For the last few years, most AI was safe by default. A chatbot summarized information. An AI assistant made recommendations. If something went wrong, it was usually just inaccurate information, which is embarrassing but not catastrophic.</p><p>Autonomous agents are different. They don’t just answer questions, they execute actions, call APIs, change data, move transactions, delete records and can turn a single instruction into a chain of actions that reshapes enterprise systems. This creates a critical gap: enterprises can deploy AI agents. But most can’t answer the basic accountability question: “Who is responsible when this agent acts?”</p><p>In a traditional system, the answer is clear. A human makes a decision. The human is accountable. With autonomous agents, the chain is different:</p><p>User request → Agent reasoning → Data retrieval → Tool selection → System change → Business outcome</p><p>Someone authorized it. Someone approved it. Someone should be able to explain what happened. But in many deployments, that someone is unclear. This gap is where incidents happen. Where compliance breaks. Where boards get uncomfortable.</p><h3>Three Concepts That Sound Similar But Mean Different Things</h3><p>Before we talk about solutions, clarify three related but distinct concepts:</p><ol><li><strong>Governance</strong> is about defining policies, roles, and processes. It answers: How should AI be controlled? What rules apply? What are the approval processes?</li><li><strong>Accountability</strong> is about assigning clear ownership and creating evidence. It answers: Who is responsible for this decision? Can we prove what happened? Can we trace each action back to authorization and policy?</li><li><strong>Autonomy controls</strong> are about runtime enforcement. They answer: What can the agent actually do? What actions are allowed? What triggers require human approval?</li></ol><p>They work together. Governance defines the rules. Accountability assigns responsibility and creates evidence. Autonomy controls enforce the rules in real time.</p><p>Most enterprises focus on governance. They write policies. They define processes. But they skip accountability. They don’t assign ownership before deployment. They don’t build evidence into the system. They discover these gaps after an incident. The enterprises avoiding incidents do the opposite. They start with accountability.</p><h3>The Five Questions That Accountability Must Answer</h3><p>An accountability framework should let you answer these questions about any agent action:</p><ul><li>Who authorized this action? (Is there a documented approval trail?)</li><li>What was the agent permitted to do? (Does the action stay within defined boundaries?)</li><li>Which data, tools, and policies influenced the decision? (Can we trace the reasoning?)</li><li>What actually happened when the action was executed? (Do we have a complete audit log?)</li><li>Who can intervene, stop, or reverse it? (Can we recover if something went wrong?)</li></ul><p>If you can’t answer all five, you have an accountability gap.</p><h3>The Control Framework That Scales With Risk</h3><p>Not all agents should have the same controls. A conversational AI answering questions is different from an agent that executes financial transactions. Smart enterprises scale controls based on autonomy level:</p><ol><li><strong>Observe level</strong>: The agent reads information and summarizes. Control focuses on access (who sees what data?). You need logging but minimal approval processes.</li><li><strong>Advise level</strong>: The agent recommends an action. Control focuses on ensuring humans review before deciding. This requires human review but limited automation.</li><li><strong>Approve level</strong>: The agent executes actions that humans have already approved. Control focuses on authorization (is the user allowed to do this?) and policy checks (does this follow organizational rules?).</li><li><strong>Autonomous level</strong>: The agent executes defined workflows without human review for every decision. Control becomes critical. You need guardrails, monitoring, auditability, and ways to recover if something goes wrong.</li></ol><p>The principle is simple: More autonomy requires more explicit controls. Trust doesn’t replace controls. Trust and controls work together.</p><h3>The Accountability Stack: What Should Be Controlled</h3><p>A production <a href="https://www.tothenew.com/insights/article/responsible-ai-checklist">AI governance framework</a> should create a clear chain from identity to outcome. This requires seven interconnected layers:</p><ol><li><strong>Identity layer</strong>: Who is making the request? Who is the agent? The system should preserve the relationship between the user and the agent, so you can always trace decisions back to specific people and systems.</li><li><strong>Authority layer</strong>: What is the agent allowed to access? What tools can it use? What data can it read? Agencies should operate on a principle of least privilege: they get only what they need for their specific task.</li><li><strong>Autonomy layer</strong>: How independent is the agent? Does it need human approval for each action? Can it make certain decisions automatically while requiring review for others? Match autonomy to risk.</li><li><strong>Enforcement layer</strong>: How do you prevent unauthorized actions? This isn’t just about the AI model. It’s about deterministic controls outside the model. Authorization checks. Transaction limits. Business rule validation. If the agent proposes an action, deterministic systems should validate it before execution.</li><li><strong>Oversight layer</strong>: What requires human judgment? High-impact decisions. Irreversible actions. Transactions above a threshold. These should trigger human review before the agent proceeds.</li><li><strong>Evidence layer</strong>: What do you record? Every agent action should create an audit trail linking the user, agent, model, policy, tool calls, approvals, and resulting changes. If something goes wrong, you can investigate.</li><li><strong>Recovery layer</strong>: How do you stop or reverse unauthorized actions? This means circuit breakers (if something looks wrong, stop). Credential revocation (disable an agent immediately). Rollback capabilities (undo transactions). Compensating transactions (if you can’t undo it, fix the impact).</li></ol><p>Together, these layers create accountability: the ability to prove that an agent acted within boundaries and did what it was supposed to do.</p><h3>How This Actually Works at Runtime</h3><p>Think about a practical example: a sales agent that recommends discounts. Discounts below 10% can be applied automatically. Discounts between 10% and 20% require manager approval. Discounts above 20% require VP approval.</p><p>Here’s the accountability flow:</p><p>Sales rep makes a request → Agent evaluates → Agent determines a 15% discount is appropriate → System checks: is this user authorized to request discounts? → System checks: does 15% require approval? → System routes to manager for approval → Manager approves or rejects → If approved, agent executes → System records: who requested, who approved, what policy applied, what changed, when it happened.</p><p>Now if something goes wrong, you can investigate. You can see exactly what happened. You can explain it to compliance. You can trace the decision chain.</p><p>Most enterprises don’t do this. They deploy the agent. The agent applies the discount. Later, someone notices inconsistencies and asks: “Why did we give this customer that discount?” Nobody can explain it.</p><h3>Building Accountability Into Agent Design</h3><p>Enterprises that scale autonomous agents successfully do something critical: they make accountability decisions before deployment.</p><p>They ask:</p><ul><li>Which business owner is responsible for this agent? (Name them.)</li><li>What data and tools does it need? (List them.)</li><li>What actions require human approval? (Define thresholds.)</li><li>How will we monitor for policy violations? (Specify metrics.)</li><li>How will we detect and reverse unauthorized actions? (Plan recovery.)</li><li>What compliance requirements apply? (Document them.)</li></ul><p>When they deploy the agent, this accountability framework is built into the system. It’s not a policy document. It’s code. It’s in the authorization layer. It’s in the approval workflow. It’s in the monitoring system. This is harder than ad-hoc deployment. It’s also the only way accountability actually works at scale.</p><h3>The Lifecycle View: Governance Doesn’t End at Deployment</h3><p>Most enterprises treat agent governance as a deployment question. “Is this agent safe to release?”</p><p>Forward-thinking organizations treat it as a lifecycle question:</p><p>Discover (What agents do we have?) → Register (Where are they deployed?) → Classify (How risky are they?) → Assign owner (Who’s responsible?) → Approve (Did we review it?) → Deploy (Is it live?) → Monitor (Is it behaving correctly?) → Review (Should we change it?) → Modify (Update policies or capabilities) → Retire (Turn it off when needed)</p><p>Each stage has governance questions. Each stage should create evidence. By the end, the organization has an inventory of agents with clear ownership, documented risk assessment, approval trails, and operational monitoring.</p><p>Without this lifecycle view, agents proliferate. Ownership becomes unclear. Monitoring is inconsistent. Governance breaks.</p><h3>Who Actually Owns Accountability</h3><p>Here’s the hard part of accountability: it requires clear assignment of responsibility before something goes wrong.</p><p>Different stakeholders own different pieces:</p><ul><li>The business owner approves the use case and accepts business risk.</li><li>Product and risk teams classify the risk level and determine what controls are needed.</li><li>Security and engineering teams define what the agent can access and what actions are permitted.</li><li>Compliance and legal teams determine what approvals are required.</li><li>Engineering and operations teams implement monitoring and incident response.</li><li>The governance function oversees the entire process.</li></ul><p>If accountability isn’t explicitly assigned to specific people with specific responsibilities, it defaults to being everyone’s job, which means it’s nobody’s job.</p><p>The best practice: create a named owner for every consequential AI agent before it deploys. Give that owner authority and hold them accountable for oversight.</p><h3>The Strategic Accountability Question</h3><p>Here’s the board-level question that separates confident enterprises from anxious ones:</p><p>“If our AI agent makes a significant decision, can we immediately answer: Who authorized this? What rules applied? What data influenced it? What actually happened? And can we reverse it if needed?” If the answer is yes to all five, your enterprise is building accountability properly.</p><p>If you hesitate on any of them, you have work to do. This isn’t about preventing all mistakes. Humans make mistakes too. It’s about visibility. It’s about being able to investigate. It’s about demonstrating that the system operated within defined boundaries.</p><p>It’s about moving accountability from aspiration to practice.</p><h3>The Path Forward</h3><p>Enterprises deploying autonomous agents in 2026 that will still trust their own systems in 2027 are doing one thing consistently: they’re building accountability into architecture and operations, not bolting it on after deployment.</p><p>This means:</p><ul><li>Designing agents with identity and authorization controls from day one, not adding them later.</li><li>Implementing runtime policy enforcement in code and automation, not just documentation.</li><li>Establishing monitoring and audit trails that create evidence of what happened.</li><li>Planning incident response and recovery procedures before you need them.</li><li>Assigning clear ownership and responsibility for each agent before it deploys.</li></ul><p>This is harder than quick deployment. It’s also the only way autonomous agents scale safely in enterprises where decisions matter. The enterprises that figure this out will <a href="https://www.tothenew.com/insights/podcast/building-ai-agent-easy-running-one-production-not">operate AI agents</a> with confidence. The ones that don’t will eventually face the question they can’t answer: “Who authorized this, and can you prove what happened?”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=07450876a211" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>