<?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 Pushpavalli Annamalai on Medium]]></title>
        <description><![CDATA[Stories by Pushpavalli Annamalai on Medium]]></description>
        <link>https://medium.com/@annamalaipushpavalli?source=rss-ed0c032a748c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*qvwGLDa5WwWk4rIJ</url>
            <title>Stories by Pushpavalli Annamalai on Medium</title>
            <link>https://medium.com/@annamalaipushpavalli?source=rss-ed0c032a748c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 11:55:48 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@annamalaipushpavalli/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[How to build an ideal accounts payable approval process?]]></title>
            <link>https://medium.com/@annamalaipushpavalli/how-to-build-an-ideal-accounts-payable-approval-process-2fe18d7231e0?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/2fe18d7231e0</guid>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Wed, 06 May 2026 10:53:09 GMT</pubDate>
            <atom:updated>2026-05-06T10:53:09.837Z</atom:updated>
            <content:encoded><![CDATA[<p>How do companies really make sure their bills get paid — correctly, availing all benefits, on time and actually improve cash flow management of a company?</p><p>The ‘Accounts Payable approval process’ makes it all happen — but the current methods of completing the process is , when put simply, a big headache.</p><p><strong>What does this headache entail, if you ask?</strong></p><ul><li>Manual data entry errors</li><li>Missed discounts</li><li>Inability to spot duplicate invoices</li><li>GST reconciliation issues</li><li>Errors in TDS calculation</li><li>Invoices arrive on different mediums — email, paper, whatsapp</li><li>PO-Invoice-GRN matching</li><li>Delays in invoice approval</li><li>Inaccurate reporting</li><li>Approval of invoices that have missing information</li></ul><p>This article summaries my learnings from talking to several financial level heads and decision makers at large Indian enterprises on— Why AP matters, challenges as of today and how to streamline it with modern day solutions.</p><p><strong>What is the Accounts payable approval process?</strong></p><ul><li>Receipt of the invoice — As soon as the good/service is delivered, the buyer generates a Goods Receipt Note (GRN). Subsequently, the supplier prepares an invoice — this elaborates in detail the quantity, unit cost, total cost and delivery specifics. The invoices are sent to the accounts team.</li><li>Invoice matching and validation — Invoices are cross checked against PO and GRN through 3-way matching mechanism to confirm that payment is only for received goods or services. The possibility for discrepancies at this stag is high. Solving for the discrepancies is very time consuming for the accounts teams.</li><li>Review and approve — Once the three way match is complete, the accounts team verifies the data. Issues are flagged, sent back to the supplier for corrections. Approved invoices go for payment.</li><li>Invoices are created on the accounting system — All the approved invoices, are manually typed in the accounting system — this creates a clear audit trail for tracking and compliance.</li><li>Vendor Payment — the finalised, approved invoices — progress to payment processing.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*l3P8Nth6aunYqjgZYx8nAQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8wc_HGhD1EWOQhLd5DuF7A.png" /><figcaption>Entire AP process, as it happens today</figcaption></figure><p>Each stage of the Accounts payable process has a distinct purpose for each role and function.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BCxfUSj4bwavI4oTuLNbjA.png" /><figcaption>Each stage of the Accounts payable process has a distinct purpose for each role and function</figcaption></figure><p><strong>How does strategic invoice management really help the underlying business?</strong></p><ul><li>Prioritizing invoice payments based on terms and early payment discounts helps <strong>optimize cash flow and strengthen vendor relationships.</strong></li><li>Regularly <strong>reconciling bank accounts</strong> ensures every transaction aligns with your records. This step helps identify discrepancies — like duplicate payments or missing funds before they spiral into costly errors</li><li><strong>Fraud prevention</strong> starts with robust AP controls — this can mean restricting system access and segregating duties. This could prevent potential misuse of authority and limits exposure to sensitive financial information.</li><li>Automating invoice matching helps flag <strong>duplicate or suspicious entries</strong>, ensuring payments only go out once.</li><li><strong>Invoice approval automation </strong>can route documents to the right approvers and send reminders to <strong>reduce delays</strong>. This eliminates bottlenecks from manual follow-ups and missing paperwork.</li><li>Tracking KPIs for Accounts payable, like <strong>invoice cycle time and processing costs</strong>, <strong>reveals inefficiencies and creates opportunities for continuous improvement.</strong></li></ul><p><strong>What are the challenges that exist in AP today and how should they be tackled?</strong></p><p>Manual AP workflows face these issues repeatedly:</p><ul><li>Manual data entry errors</li><li>Delays in invoice approval</li><li>Inaccurate reporting</li><li>Missed discounts</li><li>Inability to spot duplicate invoices</li><li>Invoices arrive on different mediums — email, paper, whatsapp</li></ul><p>Improving your accounts payable approval process means <strong>building a system that works under pressure, scales with your business, and holds up to audit scrutiny.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UYxePFmxOUTdQSJWVBMuNQ.png" /></figure><p><strong>What can an automated AP system mean for your business?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iFZxcPMXlfa1Cg33qnSqng.png" /></figure><p>As we build Elimai’s own AI enabled, GST integrated-AP spend management product for Indian enterprises, we’d love to hear your thoughts on what you believe is ‘need-of-the-hour’ when it comes to managing your business’s spend.</p><p>If you’d like to sign up for the beta version of the Elimai.spend (See every rupee. Stop every revenue leak) product— go ahead and register <a href="https://www.elimai.ai/beta-usage">here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2fe18d7231e0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Supply Chain visibility]]></title>
            <link>https://medium.com/@annamalaipushpavalli/supply-chain-visibility-af7967cbe5af?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/af7967cbe5af</guid>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Sun, 07 Jul 2024 19:49:43 GMT</pubDate>
            <atom:updated>2024-07-07T19:51:24.812Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Supply Chain Visibility</strong></p><p>The two major lessons that the world had learnt as the semiconductor shortage hit the automotive industry late in 2020 and they are as follows:</p><p>- Lack of upstream supply chain visibility issues — Most firms do not know their suppliers beyond those in Tier-1 and makes them prone to supply chain shocks in the upstream supply chain.</p><p>- Interconnectedness of global supply chains — Global Supply chains are mostly connected to one another and are far from being isolated. For example, during the supply chain crisis of semiconductors, several industries such as automotives and consumer electronics were vying for chips from the same suppliers.</p><p><strong>Why is it so hard to map Supply Chains?</strong></p><p>The resilience of a supply chain is correlated to the strength &amp; reliability of its suppliers. Yet, most lower tier suppliers are not noticed by the executives of parent companies that sit several levels above. This is especially true for OEM manufacturers that source multiple components from several suppliers. Between 5 chipmakers and 7 automakers, there is an average of 90.5 intermediary firms every year. The complex nature of these supply chains call for greater supply chain visibility.</p><p>Secondly, the supply chains are very dynamic by nature. There are only 14 intermediary supply chain firms within the semiconductor and automaker, that are common between 2003 to 2017.</p><p>Thirdly, to map an entire upstream supply chain is complex, time consuming and even impractical (given that a single firm unilaterally takes the effort). If a downstream firm attempts to do the same, it will need considerable time &amp; resources to be successful (Eg: Apple’s effort in mapping its entire supply chain). The constantly evolving nature of the supply chain also requires that a downstream firm keep its databases updated and relevant.</p><p>As manufacturing is becoming more complex with time, there is a requirement for a new set of components from new sets of suppliers. As a result, supply chains have become more interconnected, making them prone to an ever-growing threat of supply chain shocks.</p><p>Yet, when we assess the current state of supply chain visibility, we understand that most firms are focused on managing their personal buyer-supplier relationships as if they existed in complete isolation, ignoring that other firms may relate to the same suppliers or buyers. How could this change?</p><p><strong>Global Supply networks are fragmented into communities of firms</strong></p><p>All suppliers are not equal sources of risk. A few suppliers assume greater importance because they sit at the boundaries of communities. This means that these suppliers have a diverse industry representation of customers. Such suppliers sre called “High-betweeness firms.”</p><p>There are two ways to look at supply. The longer term approach would involve fixing a few tier-1 suppliers, made possible through a continuous process of mapping the supply chain. The second approach, more short term by nature would be to engage ‘High Betweeness firms.’ This could help customers gain information of the network. As supply chains become more complex, fragmented and loosely collected, the importance of high-betweeness firms have become imperative.</p><p><strong>How could mapping supply chains help companies win?</strong></p><p>Mapping upstream suppliers and downstream customers is the first step in helping companies forecast demand, that is informed by machine learning algorithms which utilize real-time information on buying patterns and identifying new parameters that affect demand.</p><p>Mapping supply chains could help create the much needed agility in decision making, whilst creating trust amidst suppliers creating an instant trustworthy platform that is designed to facilitate collaboration.</p><p>Well-funded technology incumbents within supply chain visibility include SAP, Infor, Oracle SCM, Project44 and Four Kites.</p><p>If you are building a supply chain technology company, I’d be delighted to connect! Please feel free to reach out at annamalaipushpavalli@gmail.com</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=af7967cbe5af" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How can ServiceNow be replaced — An Investment Thesis]]></title>
            <link>https://medium.com/@annamalaipushpavalli/how-can-servicenow-be-replaced-an-investment-thesis-33f3e468decc?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/33f3e468decc</guid>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[replacement]]></category>
            <category><![CDATA[servicenow]]></category>
            <category><![CDATA[startup]]></category>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Mon, 12 Feb 2024 12:56:50 GMT</pubDate>
            <atom:updated>2024-02-26T01:06:48.676Z</atom:updated>
            <content:encoded><![CDATA[<h3>How can ServiceNow be replaced?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*5rNVqWOc5cnAq5V6" /></figure><p>A market leader with a wide moat. Interweave this advantage with — phenomenal customer retention rates, high margins and an enterprise tech environment that commemorates consolidation. You land at ServiceNow.</p><p>Very simply put, ServiceNow helps in automating IT workflows. The software’s key value proposition can be concentrated into 4 critical categories — IT service management, IT operations management, IT asset management and Configuration Management Database.</p><p>The icing on the cake is that ServiceNow today isn’t just restricting its growth to being an ITSM leader but has created a footprint within managing employee &amp; customer workflows as well as enabling low-code environments for developers.</p><p>Let’s break down the factors behind why ServiceNow is successful today — (1) Key products, (2) business model (3) GTM Strategy, and (4) Market Trends.</p><p><strong>Breaking down ServiceNow’s Key Product Offerings</strong></p><p><strong>IT Service Management (ITSM)</strong></p><p>‘Work from Home’ and ‘New Normal’ are terms we are familiar with today. These new work choices, among several other factors, have further immersed us into a need for greater ‘digitization.’The monetary value of this need has increased IT spend by a staggering 69%, within a mere 10 years.</p><p>Let this not overlook the technical challenges that arise with an increasing need for IT services. Several issues can come up — technical incidents, new service requests to accommodate for, etc. Additionally, given that each request can consist of configuration of different components — (eg:) software hardware, databases, servers, etc. — how are all these different components going to communicate with each other? Is there a possibility to analyze the different incidents that may come up in the future and build technology to solve it beforehand?</p><p>Yes, this can be made possible through IT Service Management (ITSM) software of ServiceNow. It’s ITSM solution enables a company to set up an IT service desk, an incident management system, a generative AI powered chatbot (NOW Assist) for support and sealing the deal is its analytics around the support systems.</p><p><strong>IT Operations Management (ITOM)</strong></p><p>ITOM helps in storing data in different datacenters, cloud workloads and databases. ITOM also provides analytics &amp; service mapping features to help the user understand how different IT systems in an organization are intertwined with each other. This helps troubleshoot errors whenever they may arise.</p><p><strong>IT Asset Management</strong></p><p>IT Asset management refers to centralizing all IT assets of an organization in one place. This could be Enterprise Asset management (gives intel on when to replace assets), hardware asset management (computers, servers, printers, etc.), software asset management (Cloud assets, SaaS tools and databases) and Cloud optimization tools.</p><p><strong>Configuration Management Database (CMDB)</strong></p><p>CMDB manages the backend infrastructure and data foundation that supports ServiceNow. By tracking different IT assets (both hardware and software), the CMDB help categorize data buckets and create taxonomy which comprises of critical information such as the services, age, models, versions, physical locations, users and service health.</p><p><strong>Now Platform</strong></p><p>Now Platform is an Application-as-a-Platform service. It is a single data model built on a flexible data schema that delivers solutions for Anomaly detection, predictive modeling, peer benchmarking and performance forecasting. The single data model enables eliminating data silos and elevating data workflows.</p><p><strong>Service Now App Engine</strong></p><p>ServiceNow’s App Engine studio is a low-code tool for developing web applications to store information and automate business processes. The Service Now App Engine is predominantly used for: (1) Enabling users to create customized applications that optimize workflows, (2) Enhancing service delivery (3) fasten the pace of implementing digital transformation initiatives and (4) Creating reusable components such as templates and widgets that can be used across multiple applications.</p><p><strong>ServiceNow Customer Service Management</strong></p><p>ServiceNow Customer Service Management is a suite of several tools that have use cases in the following areas: (1) Single interface for agents to complete all tasks, (2) automatically assign work items to agents based on their availability, capacity, and skills, (3) Create and route cases, manage customer information and interactions, monitor case activity, (4) Using machine learning to predict fields during case creation, such as the category, priority, and assignment group.</p><p><strong>How is Service Now looking at Generative AI?</strong></p><p>Generative AI — ServiceNow has joined hands with NVIDIA to develop custom-built GenerativeAI models for different enterprises. Broadly, ServiceNow has introduced two generative AI capabilities — case summarization and text-to-code. These capabilities, fueled by the vendor’s Large Language Models (LLMs), are tailored to enhance ServiceNow’s ‘Now Platform’. This platform caters to IT operations, business management, and process automation. As for the Case summarization feature — Within the IT Service Desk, Generative AI can play a role in summarizing previous client requests and conversations to ensure smooth continuity.</p><p>In summary, for consumers of ServiceNow’s processes, services and products -&gt; technical complexities of different departments interacting with each other is hidden. This can be attributed to integrated workings around:</p><ol><li>Ability to work with multiple different channels of customer/employee engagement (including Mobile, Web and Conversational platforms)</li><li>Integrating different workflows with the signature ‘Now’ Platform</li><li>Facilitating many application integrations across different business functions</li><li>Automation of certain routine workflows saving time for different departments</li><li>Process Mining tools — Gives you real-time updates on how the processes are working, where there are bottlenecks, what is the volume and what are the variations and exceptions to that process</li></ol><p><strong>Breaking down ServiceNow’s Business Model</strong></p><p>As per the company’s public financial reports, ServiceNow has a 98% retention rate and gross profits at a whooping 78%, with the average value of ACVs priced at $42,000.</p><p>95% of ServiceNow’s revenues can be attributed to subscriptions and the rest to services.</p><p>The next question that arises would be — which verticals within ServiceNow are most successful and how much is each vertical contributing to Top-line revenue?</p><ul><li>58% of ServiceNow’s revenues rise from its IT workflows (Comprising of IT Service, Ops, Business &amp; Asset Management as well as DevOps, SecOps and Risk Management).</li><li>The second highest revenue generator for the firm is a combination of Employee &amp; customer workflows contributing to 26% of total revenues of the company</li></ul><p><strong>Breaking down ServiceNow’s GTM Strategy</strong></p><p>85% of Fortune 500 companies use ServiceNow’s products and services. With high annual contract values (~$42000/year), the software becomes affordable and savvy for use amongst the largest companies in the world. Additionally, only ~30–40% of ServiceNow’s revenues are driven through distributors, making margins very high (~78% gross margins).</p><p>With current Fortune 500 markets being saturated, ServiceNow is attempting to move down market with lower ACVs, but finds itself competing deep in the trenches with Atlassian (Parent company of Jira) where ACVs are currently ~$13,000, making the product affordable to several down market players. JIRA’s apt provisions and requirements for the down market make it a competitive contender to ServiceNow and a difficult prospect to overtake for gaining market share.</p><p><strong>Breaking down certain Market trends pushing ServiceNow’s dominance</strong></p><p>There are certain market factors that are driving the success behind ServiceNow. This includes</p><ol><li>The move towards consolidation within Enterprise SaaS (due to several factors such as down spiraling economic times, need for only a minimal number software solutions to cut on costs and do away with complexity, etc.)</li><li>The growing market and need for IT Service Management — The growth of ServiceNow over the years can be correlated to the growth of Technology in and of itself.</li><li>Incremental growth of ServiceNow through product expansion (Expanding from a pure ITSM provider to an IT infrastructure provider today)</li><li>Desire in moving down market</li><li>ServiceNow’s key feature — its ability to connect different departments of an enterprise through powerful digital workflows, makes it immensely powerful.</li><li>Service Now’s 360 degree integrated systems — current software requirements at an enterprise can be categorized into 4 major categories — IT (~79 applications) , Employee (~43 applications) , Customer (~93 applications) and Creator Workflows (~23 applications). Under these workflows, ServiceNow offers ~240 different applications for different users within the organization.</li></ol><p><strong>If we were to disrupt ServiceNow…</strong></p><p>The strategy I would like to adopt in an attempt to disrupt ServiceNow is to examine parts of the highest revenue driving vertical that users are still unsatisfied with in terms of — Product, Business Model and existing competition.</p><p>The vertical that contributes the most to ServiceNow’s key offering is its IT Workflows that comprise of — IT Service Management, Ops, Business &amp; Asset Management as well as DevOps, SecOps and Risk Management. 58% of ServiceNow’s top-line revenue is driven by this vertical.</p><p><strong>What parts of ServiceNow’s most critical offerings are customers still not satisfied with? — Product</strong></p><p>Upon analyzing hundreds of reviews on different platforms about what parts of ServiceNow is most disliked/most redundant by users, we learn that the most disliked aspects of the software are as follows:</p><ol><li>Configuration &amp; Personnel Training — Operational Issues such as high cost, complexity &amp; large resource requirement behind configuration of the ServiceNow software into current systems can make ServiceNow a drain on the enterprise’s resources</li><li>Customization — The horizontal nature, vast functionality and lack of customization choices of ServiceNow’s software makes it a broad application that doesn’t fit the needs of an organization precisely. There needs to be several customized tools to be built to make the application adaptive to the organization’s requirements. Integrating these customized tools within the existing product to make it appropriate and standardized for a particular function is currently difficult.</li><li>Complex UI — ServiceNow is an ideal application for an enterprise-level customer, but there are difficulties in day-to-day usage amongst executives in the mid-market and SMB categories. With complex user interfaces and an overload of features, ServiceNow becomes a complex tool to navigate. With hiring of sound training professionals being expensive, companies in the mid-market and SMB categories do not find ServiceNow to be a go-to IT Helpdesk Automation solution. A major drawback that is consistently relayed by several users of the App Engine application is that the UI for the ServiceNow is not very intuitive. Static dashboards are cumbersome and difficult to interpret: Dynamic Dashboards with Generative AI capabilities are required to help in increasing the ease of navigating through the dashboards.</li><li>Vertical Applications which have workflow management softwares driven to solve for the needs of particular verticals have become essential. ServiceNow has realized this need as well and introduced solutions catered to specific verticals — ServiceNow Legal Delivery, ServiceNow Financial Service Operations, ServiceNow Manufacturing Connected Workforce, ServiceNow Telecommunications Service Management.</li></ol><p><strong>What parts of ServiceNow’s most critical offerings are customers still not satisfied with? — Business Model</strong></p><ol><li>ServiceNow is an expensive software to license, given the high ACVs. There is an opportunity to build services similar to ServiceNow for down and mid-market companies (Competitors such as BMC Remedy, Atlassian’s JIRA exist but haven’t captured the entire market)</li><li>Given the current economic climate, companies are unwilling to spend excessively on technology and are not ready to incorporate several software tools into their workflows. ServiceNow’s subscription model makes IT operations management a lot cheaper than traditional giants in the space — HP, IBM, BMC. ServiceNow also has over 1000+ customers that pay $1 million/year as ACV. Yet, given the host of solutions that ServiceNow is very broad, several organizations may not require the vast number of tools that the software provides. They could reduce their cost, when led by a feature-based pricing model over a flat subscription fee.</li></ol><p><strong>What parts of ServiceNow’s most critical offerings are customers still not satisfied with? — Competition</strong></p><p>Unlike some of its competitors, ServiceNow has grown to become both a Platform-as-a-Service (PaaS) and Software-as-a-Service (SaaS) product. How, you may ask? PaaS emerges because of the platform’s capabilities in creation of custom applications, and SaaS because it offers a suite of ready-to-use software applications accessible over the internet ‘as a service’.</p><p>The upmarket is saturated (85% of fortune 500 companies already use ServiceNow).</p><p>Atlassian’s JIRA, which currently serves the ITSM needs for the down market is actively looking to move upmarket. Annual growth of the initial ACV is decreasing by the year. This implies that the value created in its new offerings are not enough. To move down market and into more competitive segments such as low-code,no code — new distribution channels with partners have to be built from ground up. With companies such as Microsoft having an extraordinary distribution channel in place, if Microsoft becomes aggressive around the ITSM space, it would be extremely hard for ServiceNow to scale. With the incoming of several startups, there are several generative AI first companies that are looking to tackle the automated help desk management space that could be potential competitors to ServiceNow in the near future as well.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=33f3e468decc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Which APIs become Independent Companies?]]></title>
            <link>https://medium.com/@annamalaipushpavalli/which-apis-become-independent-companies-550c482f6aef?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/550c482f6aef</guid>
            <category><![CDATA[founders]]></category>
            <category><![CDATA[api]]></category>
            <category><![CDATA[venture-capital]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[investors]]></category>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Thu, 04 Jan 2024 15:24:13 GMT</pubDate>
            <atom:updated>2024-01-05T13:23:09.838Z</atom:updated>
            <content:encoded><![CDATA[<p><em>If you are building an API-first company, please feel free to </em><a href="https://www.linkedin.com/in/pushpavalli-pushpa-annamalai-1b8a7315b/"><em>reach out</em></a><em>!</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Gjkrzn7v_p8VYGjt.png" /><figcaption>Source: Grace Isford, Lux Capital</figcaption></figure><p>APIs are not new. There are plenty. Yet, a few APIs tend to be large enough to become independent companies that went public at billion dollar valuations. Does the API you are building right now, have the same potential? In this post, I explore the history behind how APIs came into being and share a 5-part framework of how to assess the potential of your API by dissecting past winners in the space.</p><p><strong>What is an API?</strong></p><p><a href="https://posts.technically.dev/archive/whats-an-api">Application Programming Interface (APIs)</a> is a way for two or more <a href="https://en.wikipedia.org/wiki/Computer_program">computer programs</a> to communicate with each other. APIs are a set of subroutine definitions, communication protocol tools for building software.</p><p>Let’s put this simply: Imagine a Lego Castle built with lego pieces. Each Lego block is an API. When these lego blocks are meaningfully connected with each other, they create a complete castle. Similarly, when APIs are made to meaningfully interact with different functions, they create impact for businesses.</p><p>Examples of APIs: Weather Application on our phones: (1) Weather Bureau’s software systems have daily weather data -Weather App on our phone talks to the system via an API and shows daily weather data on our phones. (2) Uber integrates with Google’s Maps API to track drivers and Stripe’s API to process payments for rides.</p><blockquote>On an average, it is said that every modern day application uses 17 different APIs.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vZC__wVrOuoCN0Zi" /></figure><p><strong>Growth of APIs — A History</strong></p><blockquote>“By bettering each part, the whole can be bettered as never before.” — Henry Ford</blockquote><p>Henry Ford’s ‘Assembly line’ innovation was not a success purely because it was making things more efficient. That was the result of the innovation. Henry Ford’s real victory was the Assembly line combined a lot of things that were not solely manufactured in-house, to create a car.</p><blockquote>APIs follow a very similar trajectory and have caused the “Industrial Revolution of Software” by making software scalable.</blockquote><p>To get a sense of which APIs make the unicorn cut, let’s dive into History.</p><p>Imagine, computers in the 70s. Computer companies (IBM, Sun Microsystems, etc.) during this era built all components of computers in-house — Mainframes, RAMs, Chips, Software &amp; Hardware. As adoption of computers expanded, there emerged an ecosystem of several thousand companies making <strong>each component </strong>— mainframes, software and hardware (Manufactured in different places, but sold as one single final unit).</p><p>Next, with the vast variety of software required — a distinction arose. Software players started specializing in two categories- builders of operating systems and builders of applications. With time, the size of the opportunity to build large companies <strong>purely by building applications </strong>was on the rise.</p><p>In early 2000s, all technology companies had their own physical servers. Heavy, Monolithic applications were built in-house. The product stack emerged and comprised of both Infrastructure (Hosting, communications, monitoring &amp; analytics, payment, security) and Features (Features through which users interact). Late 2000s, saw the emergence of several players in the infrastructure layer, but most components still needed to be coded in-house.</p><blockquote>A very significant expansion occurred when single applications became disaggregated. Logical units within applications (functions) started becoming independent companies. The interface to the different functions is an API.</blockquote><p>The API explosion happened in two different directions — Horizontal (Common APIs for a large number of Industries) &amp; Vertical (APIs for specific Industries). Horizontal APIs emerged as a result of many components becoming ‘API-fied’. Whereas, Vertical APIs emerged as more APIs started unbundling product features. This can be explained below:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/964/1*huZdWhPG9Xe4pqqoj_8i9Q.png" /></figure><blockquote>“Every time there is a disaggregation/unbundling of a traditional technology product, the market tends to grow by a safe factor of 10 factorial.”</blockquote><p>Today, there are several API-first companies that are large, publicly listed unicorns. These companies include <a href="https://stripe.com/">Stripe</a>, <a href="https://www.twilio.com/en-us">Twilio</a>, <a href="https://goshippo.com/">Shippo</a> to name a few.</p><p>What led to APIs becoming as massive as they did? This could be attributed to Technology &amp; markets maturing at unprecedented levels. Markets were maturing because more and more products were being built. Technology matured since the Markets of customers needing Feature-as-a-Service shot up + number of Infrastructure APIs were also going up.</p><blockquote>Better Infrastructure APIs = easier to build ‘feature-as-a-service’ APIs.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*tfqabgI-JNn8cd_-.png" /><figcaption>Source: Point Nine Land</figcaption></figure><p><strong>How are APIs different from SAAS products?</strong></p><p>APIs are inherently the building blocks of SAAS products.</p><blockquote>APIs solve for smaller, yet hard to crack problems for SAAS products (Eg: Payment processing or identity verification).</blockquote><p>Yet, the differences between API and SAAS are as follows:</p><p><strong>(1) Built for the Developer </strong>— One needs to be able to write code to use an API, whereas software has a User interface that can be used by both technical &amp; non-technical users.</p><p><strong>(2) Seamless unnoticeable integration into consumer applications — </strong>When we use any consumer application, we would hardly notice that an API is built into it. The processes are woven together seamlessly, making the consumer unaware of the existence of an API. SAAS, on the other hand, works as the user interface.</p><p><strong>(3) Modular</strong> — Developers can use a few/all endpoints of APIs, depending on their needs. Your first interaction with the service is how your company is perceived. For eg: In Stripe, you first interact with an endpoint as a developer. Therefore, Stripe is considered to be an API-first company.</p><p><strong>(4) Code is the primary User Interface </strong>— Developers build dedicated Graphical User Interfaces (GUIs) on top of these APIs — Which would be customized for their own applications. Eg: Stripe (A payment processing API) has an E-Commerce Application built on top of it, to process payments.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*szRO009Z9VVpxHzc.png" /><figcaption>Source: Point Nine Land</figcaption></figure><p><strong>Power of APIs</strong></p><blockquote>API makes: (1) Software easier to write than ever before, (2) Barriers to creating a tech startup drops significantly since every startup can focus on its own core competency and outsource other functions to other API-first companies instead of building all components in-house.</blockquote><p>The beauty is APIs are not just revolutionizing software, but there are also eating out traditional businesses such as Telecom, Banking &amp; Healthcare.</p><p><strong>What APIs can really become companies?</strong></p><p>A good candidate to become an API company is:</p><ul><li><strong>Simplifies Complexity: </strong>A previously existing complexity vanishes with an API. A very clear quantifiable ROI needs to exist. This should be that the API makes operations cheaper or that it saves time. Eg: <a href="https://plaid.com/products/auth/?utm_source=Google&amp;utm_medium=paid&amp;utm_campaign=DEPT_SEM_Google_Brand_ACQ_Performance_NAMER_US-CA_Core_CPA-ROAS_BAU_Exact&amp;utm_term=Core_KW-Exact_Plaid&amp;utm_content=plaid&amp;gad_source=1&amp;gclid=CjwKCAiAqNSsBhAvEiwAn_tmxZAZ65Sf-SBase3wIG3ly6Y8D_2yQZBqYQPBEAvsPvnihFfOJpopDBoCO8IQAvD_BwE">Plaid</a>, <a href="https://stripe.com/?utm_campaign=US_en_Search_Brand_Stripe_EXA-20839462206&amp;utm_medium=cpc&amp;utm_source=google&amp;ad_content=683853401230&amp;utm_term=stripe&amp;utm_matchtype=e&amp;utm_adposition=&amp;utm_device=c&amp;gclid=CjwKCAiAqNSsBhAvEiwAn_tmxZRFwiDFgAQvFkBe5DfF3yajA8SbSAx7s7ROfO_C59ApaMMIYlIfshoCU1IQAvD_BwE">Stripe</a>, <a href="https://goshippo.com/">Shippo</a>, <a href="https://www.twilio.com/en-us">Twilio</a></li><li><strong>Large TAM: </strong>The API has independent functions that are common to many applications (Eg: Payment processing/Storing Data).</li><li><strong>Business Critical Functionality: </strong>The functionality of a third-party API may not have to be a company’s core competency, but should be solving for a critical need for the business.</li><li><strong>Seismic Shift in the Technology Landscape: </strong>The most successful API companies sprung through during times when there is a significant event in technology history. Eg: (1) Twilio — Smartphones &amp; Apps on App store became huge. App builders required a telecommunication tool to verify identity. (2) Stripe-E-Commerce was on the rise with Shopify helping merchants sent up online storefronts with ease. This was coupled with the rise in usage of credit cards over cash, requiring a tool for payment processing.</li><li><strong>Community driven X-Factor:</strong> Most successful API companies have built a strong, vast community of early users mostly using open source software. Eg: Stripe started as a single repository on Github where it helped nurture its earliest developer users.</li></ul><p>More on this framework can be read <a href="https://gisford.medium.com/the-third-party-api-economy-891b2a774fa5">here</a>.</p><p>Let’s look through some API winners, through a five factor lens:</p><p><strong>Plaid, Current Valuation: $13.4 Billion</strong></p><ul><li><strong>Superpower: </strong>Payments platform that enables an individual to securely connect their financial bank accounts to other apps.</li><li><strong>Large TAM: </strong>The B2B Fintech Payments market is estimated at $125 Trillion (Gartner).</li><li><strong>Simplifies complexity: </strong>Makes it easier for 3rd party apps to get financial data while protecting the individual’s personal financial data</li><li><strong>Business Critical Functionality: </strong>Secure financial data sharing might not be a core competency for different financial apps, yet the need to share financial data securely is critical to both the bank &amp; the financial application</li><li><strong>Seismic Shift: </strong>With the mobile-first banking and the ease in payment processing, there has been a spike in online e-commerce brands and financial investment applications, which required access to your bank account details for payment transfer.</li><li><strong>X-Factor: </strong>Starting as a white-label account linking service, Plaid transitioned to a consumer brand by putting its logo on the bank linking experience inside fintech apps, creating a Visa-like 3-sided network of 500M customers, 11,000 banks, and 2600 fintechs.</li></ul><p>If you are building interesting Business software products using APIs or attempting to build an API-first company, I would be delighted to connect. Please reach out to Pushpavalli Annamalai on <a href="https://www.linkedin.com/in/pushpavalli-pushpa-annamalai-1b8a7315b/">LinkedIn</a>!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=550c482f6aef" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[FinTech x B2B Verticalization (Market Map)]]></title>
            <link>https://medium.com/@annamalaipushpavalli/fintech-x-b2b-verticalization-market-map-1ea8855444d6?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/1ea8855444d6</guid>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Mon, 13 Nov 2023 19:33:37 GMT</pubDate>
            <atom:updated>2023-11-13T19:33:37.998Z</atom:updated>
            <content:encoded><![CDATA[<p>B2B Payments is a $125 Trillion industry. The B2B Fintech Market is growing at a CAGR of 32% annually. Yet, this industry is far from being digitally disrupted.<br>Involving financial services within a vertical SaaS solution is the layer cake strategy that would make vertical software much more sticky by nature.<br>This market map illustrates some companies that are building financial services for specific verticals.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*q90NHmKlXzg4Ik4D2gODcA.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1ea8855444d6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Industry-specific SaaS x Generative AI-Native Companies]]></title>
            <link>https://medium.com/@annamalaipushpavalli/industry-specific-saas-x-generative-ai-native-companies-a1782addc173?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/a1782addc173</guid>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[healthcare-technology]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[saas]]></category>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Fri, 27 Oct 2023 22:26:43 GMT</pubDate>
            <atom:updated>2023-10-27T22:26:43.326Z</atom:updated>
            <content:encoded><![CDATA[<p>There are three consumer-driven industries that have already hit $100 Million in ARR, driven by AI-Native companies (Code writing, Copy Writing and Image Generation). More than 100 Million people have experienced the ‘GPT’ moment and the tectonic plates of technology are shifting.</p><p>Along the same lines, we are seeing the rapid growth of AI-Native companies within B2B verticals, challenging incumbents within the space. There exists a perception that enterprise software is ill equipped to handle multifaceted workflows that work with unstructured data (Eg: videos, images &amp; audio files). Large Language Models(LLMs) are shattering this notion.</p><p>What are these verticals and what are the use cases that might prove Generative AI would be a blockbuster within them?</p><p><strong>(I)</strong> <a href="https://a16z.com/where-will-ai-have-the-biggest-impact-healthcare/">Generative AI will possibly have the greatest impact within..Well, yes. Healthcare. </a>Enterprise software has had struggles in penetrating into healthcare for several reasons — (1) Healthcare systems are overburdened with several siloed IT teams, (2) burnt out staff and (3) as a sector, healthcare is inherently slow to adapt to technological advancement.</p><p>Developing novel drugs to cure life-threatening diseases is important. But, even if we were to cure all cancer in the world,<a href="https://www.washingtonpost.com/archive/politics/1990/11/02/major-longevity-gains-termed-unlikely/1eeffaef-b2fb-4e2c-b17d-3c487208f25b/"> it is expected to increase life span by only 3 years.</a> Instead, some of the bigger problems in healthcare can be cured through technology. Healthcare is (1) a data, operations &amp; logistics problem, (2) a consumer and engagement problem.</p><p>Despite these drawbacks, healthcare is also probably the most lucrative space for technology to capture, merely because of the size of the opportunity: <a href="https://a16z.com/the-biggest-company-in-the-world/">20% of US GDP ($4 Trillion) is spent on healthcare.</a></p><p>Yet, low penetration of technology within Healthcare is about to change. How?</p><p>Healthcare would bypass the Healthcare enterprise software stage to move from <a href="https://a16z.com/where-will-ai-have-the-biggest-impact-healthcare/">‘Fax Machines’ (current technology used in a few spaces within the industry, as surprising as it may sound!) to Generative AI.</a></p><p>Where is the proof? Let’s delve into use cases behind where we are seeing generative AI companies fixing gaps within the Medical ecosystem.</p><p>(1) <strong>Medical Scribes</strong> — Generative AI is propelling the narrative of ‘Let Doctors go back home on time.’ When doctors look to examine patients, providing right diagnosis and treatment is not all the work they do — preparation of meticulous notes customized to suit the requirements of 3 stakeholders, namely — other Doctors, Insurance providers and patients is an important aspect of their job. The patience and cognitive requirement to complete these clerical tasks is high and exhausts a doctor by capturing upto 20 hours of their time every week.</p><p>An AI driven automated medical scribe captures nuances of clinician-patient conversations in real time and automatically generates notes from the conversation, customized to fit the individual preferences of the reader. AI-Native companies that have emerged with the Medical scribe as their core offering include — <a href="https://www.abridge.com/">Abridge</a>, <a href="https://www.ambiencehealthcare.com/">Ambience</a>, <a href="https://www.deepscribe.ai/">Deep Scribe</a>, <a href="https://www.nabla.com/">Nabla</a>, to name a few.</p><p>(2) <strong>Intelligent Care Management </strong>— There are several categories of healthcare that require care beyond the hospital/physician’s clinic. Examples include Maternity (Fertility, Postpartum), Surgical Care (Pre/Post-Surgery Engagement), Cancer (Full-time on support), Gastroenterology(Pre/Post Colonoscopy preparation) and Chronic Care Management (Managing disease at home). Today, engagement with a patient beyond the hospital/physician’s clinic is either ineffective or not fully solved for. Generative AI can change this equation and act as an Intelligent Care provider for recovering patients at all times. Companies providing Intelligent Care Management as a core offering include — <a href="https://www.memorahealth.com/">Memora Health</a>.</p><p>(3)<strong> LLMs for diagnosis</strong> — As far fetched as this may sound, there are companies that are building Medical LLMs that are pre-trained on evidence based healthcare content and implement reinforcement learning from human feedback (RLHF), in this case from Certified Clinicians for diagnosis. Companies within this space include: <a href="https://www.hippocraticai.com/">Hippocratic AI</a>, <a href="https://glass.health/">Glass Health</a>, <a href="https://www.ycombinator.com/companies/latent">Latent</a>.</p><p>(4) <strong>AI Assistant </strong>— A clear example of an AI assistant to help you learn today would be ChatGPT. But, ChatGPT might not be an effective tool to scan through thousands of pages of Insurance policy documents to extract all the rules, possibility of claim acceptance/rejection and send/receive data to any Electronic health/Medical record (EHR/EMR) with ‘NO’ development involved. LLMs pre-trained on health data of several years might be a more appropriate fit for this use case. We see companies capitalizing on this trends through building ‘ChatGPTs for Health Plan Operations’ and AI-driven customer success teams that help in navigating complex insurance rules within minutes. Companies building in this space include: <a href="https://www.co-helm.com/">Co-Helm</a>.</p><p>(5) <strong>AI in Clinical research </strong>— Generative AI is proving to be exceptionally useful in generating evidence on demand for clinical research. AI can provide personalized evidence for every clinical decision and research inquiry. <a href="https://wiki.pathmind.com/generative-adversarial-network-gan">Generative Adversarial Networks (GANs)</a> or Variational Autoencoders are being used for this purpose. Without the use of technology, it takes months to collect medical data and transform it to real world evidence — decelerating the pace of research, and de-optimizing clinical trials. Researchers rely on iterative experimentation — testing to create the ideal combination of elements, proportions, etc. Today, it is possible to go from Idea to publication within a few weeks and the process can augment staff efficiency as well. Companies building in this space include: <a href="https://www.atroposhealth.com/">Atropas Health</a>.</p><p>(6)<strong> Revenue Cycle Management</strong> — It takes days on end to process every claim in a manner that avoids denials from insurance companies. There exists a pain point in making sure patients settle their bills on time. Generative AI applications are being built to solve for these critical pain points. Companies building in this space include: <a href="https://www.athelas.com/">Athelas</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1018/1*BBGa2YjAoNe-xYKr4RY2HQ.png" /><figcaption>Source: Sequoia Capital</figcaption></figure><p><strong>(II) </strong>The second sector that Generative AI is swinging up is — Legal Technology.</p><p><strong>(1) AI for Injury Lawyers</strong> — Generative AI can help turn medical documents (medical reports, treatment records) and legal case files (accident reports, witness statements) to create demand packages that will help injury lawyers prepare for legal proceedings.</p><p><strong>(2) Contract Analysis — </strong>Generative AI within legal technology has a massive impact in analyzing contracts, flagging areas that need review and provides suggestions on how to negotiate based on legally approved guidelines. Additionally we are seeing AI being helpful with regulatory compliance management &amp; due diligence as well. Legal teams can also build contract playbooks using Generative AI and apply them within MS word. Companies building in this space include: <a href="https://www.ycombinator.com/companies/pincites">Pincite</a>, <a href="https://www.harvey.ai/">Harvey AI</a>, <a href="https://casetext.com/">CaseText</a> and <a href="https://ironcladapp.com/">IronClad</a> (Incumbent with new technology).</p><p><strong>(3) Acting as Co-Counsel (AI Assistants) — </strong>Generative AI’s most common use case is its ability to act as an Assistant and this is no different in the legal tech space. Co-Counsels as they are popularly referred to, help lawyers prepare document reviews, legal research memos and deposition preparations. Companies building within this space include: <a href="https://www.ycombinator.com/launches/J2U-pincites-ai-powered-contract-negotiation">Pincite</a>, <a href="https://www.harvey.ai/">Harvey AI</a> and <a href="https://casetext.com/">Case Text</a></p><p>(III) Fintech is another space that Generative AI seems to have emerging scope in. Accuracy has always been an issue around LLMs due to hallucinations, and given its core requirement in finance, one would not presume that generative AI would play a massive role here. Despite these restrictions, Generative AI has found niche buckets within FinTech where they can be very useful. These include:</p><p><strong>(1) Fraud Detection: </strong>There are 3 terms that one must be familiar with while thinking about Generative AI — Autoencoders (AI systems to compress data), Large Language models (Eg: ChatGPT) and <a href="https://towardsai.net/p/l/generative-ai-gans">Generative Adversarial Network</a> (GAN — Comprises of a content generator and a content discriminator).</p><p>The focus for fraud detection in Generative AI would be through GANs (generative adversarial networks). GANs can create synthetic data that mimic real world transactions. The generator creates these new transactions and the discriminator evaluates the authenticity of these transactions.</p><p>Why would creating synthetic data help?</p><p>(1) The sheer volume of synthetic and real world data can help GANs train more efficiently. The lack of data creation in traditional models may push the operating efficiency of some models far behind.</p><p>When the GAN model is trained on a dataset that comprises of both legal and illegal transactions to learn patterns, they can be used to identify suspicious activity and predict under-the-radar anomalies.</p><p><strong>(2) Company specific financial search engines</strong>: Private and public financial data is critical to every company. To retrieve information, several databases/Scanned PDFs/PowerPoints/Spreadsheets must be browsed or data should be accessed through querying using programming languages. Through LLMs, this data can be captured through natural language queries, saving man-hours and increasing efficiency. Given the sensitive nature of financial data, there is a rise in solutions that are encrypted end-to-end. Companies building in this space include: <a href="https://www.hebbia.ai/">Hebbia</a>.</p><p>If you are building Generative AI-driven applications within traditional industries, I’d be delighted to connect. Please find me on <a href="https://www.linkedin.com/in/pushpavalli-pushpa-annamalai-1b8a7315b/">LinkedIn</a> or email me at annamalaipushpavalli@gmail.com!</p><p><em>Please note that the list mentioned above is not exhaustive. These are opinions expressed by a single individual. Sole discretion is advised for </em><strong><em>all</em></strong><em> purposes.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a1782addc173" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Generative AI != Traditional Software A Take on evolving Business Models in Technology]]></title>
            <link>https://medium.com/@annamalaipushpavalli/generative-ai-traditional-software-a-take-on-evolving-business-models-in-technology-a5205d9f73a6?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/a5205d9f73a6</guid>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Sun, 27 Aug 2023 22:04:48 GMT</pubDate>
            <atom:updated>2023-10-02T15:25:15.362Z</atom:updated>
            <content:encoded><![CDATA[<h3>Generative AI is not Traditional Software (A Take on evolving Business Models in Technology)</h3><p>AI is pretty much seen as the Future of Software. This can possibly be attributed to how AI Applications look and feel upon usage: They have a user interface, manage data and integrate with other systems. Yet, at their core, AI applications are different. They essentially aim to <a href="https://karpathy.medium.com/software-2-0-a64152b37c35">‘Train Data models’ to help make predictions or decisions</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/758/1*fA86cKSVerrUq1DOfspJVQ.png" /><figcaption>Source: Global X ETF</figcaption></figure><p>Without a doubt, AI will transform Businesses for the better. But, this begs the next question. Would AI Applications see the same monetary success as Traditional Software, SaaS and Cloud-SaaS?</p><p>Seeing the <a href="https://www.theguardian.com/technology/2023/feb/02/chatgpt-100-million-users-open-ai-fastest-growing-app">rapid adoption of LLMs</a> and Generative AI tools in general, the answer to the previous question should be a no-brainer.</p><p>But, in reality, the answer is ‘It is too early to tell’.</p><p>Why, you may ask?</p><p>(1) AI Companies have much lower gross margins in comparison to Traditional Software Companies — Gross margins in AI Applications range between 50–60%, whereas Software attracts a Gross margin ranging between 60–80%. This difference can be attributed to <a href="https://i0.wp.com/a16z.com/wp-content/uploads/2023/04/3.-Prices-for-common-GPU-types-by-cloud-provider.png?ssl=1">extremely high compute costs</a>, <a href="https://i1.wp.com/a16z.com/wp-content/uploads/2023/04/2.-Available-GPUs-by-cloud-provider.png?ssl=1">high infrastructure costs</a> and requirement of on-going human support for AI. The success of an AI company has become highly dependent on their <a href="https://a16z.com/2023/04/27/navigating-the-high-cost-of-ai-compute/">access to compute resources</a>. The demand for compute power today exceeds its supply by a multiple of <a href="https://cset.georgetown.edu/publication/ai-and-compute/">10!</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/486/1*yTKqx03cVur0QDmqTxefqg.png" /><figcaption>Source: a16z</figcaption></figure><p>Training AI models is by no means a joke. Manual cleaning and labelling of data sets is a major barrier in large scale adoption of generative AI. The process doesn’t end with just training the model too. Once the model is deployed, new data must be captured, labelled and fed into the system. Re-training becomes a critical component to enable accuracy in results.</p><p>At the current stage of development of AI models, humans are plugged in to monitor, review and augment AI systems. Some data is processed by humans and some gets processed through computers. Both, humans and AI together become joint decision makers. With time, AI models are set to improve and human intervention would decline.</p><p>As the compute cost goes high, the on-going human support cost reduces and vice versa. This is vastly different from the traditional software business models. Traditional Softwares run on an almost <a href="https://ckluis.com/the-marginal-cost-of-software-approaches-zero-7fda166f219f">marginal zero-cost business model</a>. For every additional unit of software sold, there is additional product development cost/cost for supporting the sale only for iterative versions. Software can be sold as a ‘Product’, while AI would be seen as a ‘Software+Service’ business model. This point serves as perfect leeway into scaling.</p><p>(2) Scaling — AI models are particularly hard to scale because of the notorious ‘<a href="https://venturebeat.com/ai/living-on-the-edge-how-edge-cases-will-determine-the-future-of-generative-ai/#:~:text=In%20AI%20development%2C%20success%20or,usability%20of%20an%20AI%20tool.">Edge Cases</a>.’ Developers who train models cannot accommodate only for the most common use cases. 40–50% of the intended functionality of an AI product can reside in Edge Cases. To accommodate for Edge Cases, a vast quantity of user data must accumulated and trained. As the model matures, edge cases become increasingly expensive to accommodate. These issues invariable go back to the issue of high compute costs.</p><p>On the other hand, given that Software is sold as a ‘Product’, the hassle of scaling resides only with the distribution teams, not the Product/engineering teams.</p><p>(3) Defensibility Moats — With new model architectures being developed in open, academic settings and pre-trained models available open source, the struggle to create differentiation in AI is real. Many entrepreneurs invests heavily in data collection and labelling. But, it is important to realize that beyond a point, <a href="https://a16z.com/2019/05/09/data-network-effects-moats/">data can be an empty moat</a> and be commoditized.</p><p>Traditional software has very clear defensibility moats. They have functioned for over a decade on a simple premise: Technically superior products that reach the relevant audience quickly, create exclusivity and see major brand advantages.</p><p>Where does this leave AI?</p><p>Despite all the drawbacks, three industries using <a href="https://www.weforum.org/agenda/2023/02/generative-ai-explain-algorithms-work/">Generative AI</a> have already touched a $100 million in Annualized Revenue (Image Generation, Copywriting and Code Writing).</p><p>What’s made them successful?</p><p>AI would thrive in industries that perform high scale tasks that are minimally complex in nature. It is important to choose problem domains very narrowly to implement AI and also plan for high variable costs during its implementation. For example, companies in the CRM space have identified AI to be incredibly helpful in updating records, while Malware detection companies have utilized AI to show remarkably better performance. With time, we can see AI products that automate model training, standardize developer workflows as well as monitor and secure AI models in production.</p><p>I would love to connect with startups that are building in the Generative AI/Vertical SaaS space! Please connect with <a href="https://medium.com/u/c5a779c6b7b0">Pushpavalli Annamalai</a> on Medium/ <a href="https://www.linkedin.com/in/pushpavalli-annamalai-1b8a7315b/">LinkedIn</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a5205d9f73a6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI products are not Global products. Selling AI products Globally is still difficult.]]></title>
            <link>https://medium.com/@annamalaipushpavalli/ai-products-global-products-selling-ai-products-globally-is-still-difficult-463cf773427c?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/463cf773427c</guid>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[generative-ai-tools]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[technology]]></category>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Sat, 05 Aug 2023 17:12:22 GMT</pubDate>
            <atom:updated>2023-10-19T19:16:05.726Z</atom:updated>
            <content:encoded><![CDATA[<p>A slew of <a href="https://twitter.com/sonyatweetybird/status/1584580362339962880/photo/1">AI-Native</a> and <a href="https://bowerycap.com/blog/insights/artificial-intelligence-in-vertical-software-a-map-of-the-market">AI-Embedded Apps </a>have been built over the last few months. Thanks to the incredible <a href="https://i2.wp.com/a16z.com/wp-content/uploads/2023/08/2741_-Unprecedented-Speed-to-Adoption.png?ssl=1">rate of consumer adoption</a> of Generative AI.</p><p>The current B2C &amp; D2C e-commerce driven business models involve selling physical goods through digital channels. Given the slew of AI products in the market and the <a href="https://www.bvp.com/atlas/artificial-intelligence-is-generating-massive-consumer-surplus">attractive gains it bring to the consumer</a>, the future would entail <a href="https://bloggingwizard.com/ecommerce-platforms-sell-digital-products/">selling AI-based digital products through digital channels</a>.</p><p>There are two aspects to making a digital product successful.</p><p>Building the product and driving their adoption across the world at the quickest pace possible.</p><p>Sounds simple, doesn’t it?</p><p>Except, to drive adoption globally, companies must:</p><ol><li><strong>Comply to region-specific Tax regulations:</strong></li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AORW7-OyuVp1HWJGDhMKLA.png" /><figcaption>Source: PwC</figcaption></figure><p><strong>2. Comply to region specific Privacy Regulations</strong></p><p>GDPR(EU), California Consumer Privacy Act(United States), Australia Privacy Act, Personal Data Protection Bill and General Personal Data Protection Law (Brazil) and several other country specific privacy regulations must be adhered to. This is a significant challenge in making AI products global</p><p><strong>3. Drive Location specific Product Discovery and Marketing</strong></p><p>Marketing products through tools/applications that caters to a large audience in a particular region is a route to creating greater digital product adoption.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MrSEvwowVVmlC_CbhEeGeg.png" /><figcaption>Applications/Digital Tools facilitating Product Discovery in specific regions of the world</figcaption></figure><p>To tackle these issues, a need for a service that can <a href="https://i2.wp.com/a16z.com/wp-content/uploads/2023/07/The-Next-Shopify-Article-Inline-02-2.png?ssl=1">stitch all aspects of the Post Product-building phase</a> is imperative. This includes involvement of a Payment Service provider, Metered Billings/Subscription Management, Global Tax Calculation, Fraud Prevention and Marketing, Customer Retention and Support Activities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*g9WY-b9cZ8viWade.jpg" /></figure><p>This would allow Startup founders to outsource a large part of their scalability issues and utilize their time to focus on what they do best: Build Category-Defining Products.</p><p>Merchant of Records(MOR) is a powerful tool that can help in building product scalability at a rapid pace. As customers, we utilize MOR’s services pretty much everyday : Through <a href="https://a16z.com/2023/05/31/generative-ai-is-coming-for-insurance-may-2023-fintech-newsletter/#section--2">Apple’s App Store</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GY9NMLcgp562rhxY1ZzVMA.png" /><figcaption>App Store as a Merchant of Record</figcaption></figure><p>Merchant of Records can be a game-changer for scaling AI driven digital products. With<a href="https://i1.wp.com/a16z.com/wp-content/uploads/2023/07/The-Next-Shopify-Article-Inline-03-3.png?ssl=1"> startups and Incumbents</a> just emerging to build in the Generative AI MOR sector, the space is ripe for innovation.</p><p>I would love to talk to any startups building for this niche! Please connect <a href="https://medium.com/u/c5a779c6b7b0">Pushpavalli Annamalai</a> on Medium/LinkedIn or @@PushpaVenky99 on Twitter</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=463cf773427c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Most Effective Way for AI-Native Companies to meet unique needs of their markets]]></title>
            <link>https://medium.com/@annamalaipushpavalli/most-effective-way-for-ai-native-companies-to-meet-unique-needs-of-their-markets-engaging-users-a-120ac8b55e54?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/120ac8b55e54</guid>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[healthcare]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[genai]]></category>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Wed, 19 Jul 2023 19:50:19 GMT</pubDate>
            <atom:updated>2023-08-01T20:54:28.194Z</atom:updated>
            <content:encoded><![CDATA[<p><a href="https://www.abridge.com/">Abridge</a>, an ‘AI for medical conversations’ company, started out with no intention of leveraging AI or earning a premium in valuation with an ‘AI Startup’ tag. <a href="https://www.linkedin.com/in/shivdevrao/">Shiv Rao</a>, a cardiologist based in Pittsburg and founder of Abridge looked to merely solve a critical problem that his industry has been facing for ages: Doctor burnout.</p><p>The reasons behind doctor burnouts can be many, including but not limited to: (1) High stress levels, (2)Odd hours of work and (3)Mind boggling amounts of clerical work.</p><p>Yes. Clerical work. Surprising isn’t?</p><p>When doctors look to examine patients, providing right diagnosis and treatment is not all the work they do — preparation of meticulous notes customized to suit the requirements of 3 stakeholders, namely — other Doctors, Insurance providers and patients is an aspect of their job as important as providing diagnosis and care. The patience and cognitive requirement to complete these clerical tasks is high and exhausts a doctor as frequently as once in 15–20 minutes and diverts them from performing at the job they were initially hired for— providing the best possible care to patients and enabling them to stay at the top of their health.</p><p>Shiv Rao looked to slash out a whale’s share of this clerical work, resulting in the birth of Abridge. The team at Abridge have trained existing data models using a proprietary data set with 1.5 million medical conversations that have been collected upon patient consent for several years.</p><p>Today, the trained AI utilized by Abridge can summarize conversations between Doctors and Patients and can extract data with respect to billing, coding and risk assessment. The AI has the power to separate medically relevant parts of the conversation through its transcript monitoring feature.</p><p>Cool, isn’t it? Wait, it just gets better.</p><p>Abridge has recently built a consumer facing app to improve on the notes released by the AI. The consumer app allows patients and their families to share feedback on the AI generated notes during their conversations with doctors, leading to Abridge receiving more richer data to train their models. This ultimately ends up serving the enterprise in making our systems more intelligent. Keeping the ‘customer’ (Patient in this case) as the center of the solution and engaging them through the right tools can help AI companies effectively meet out the needs of their respective markets</p><p>By: Pushpavalli Annamalai</p><p>References:</p><p>(1) <a href="https://www.abridge.com/">https://www.abridge.com/</a></p><p>(2) <a href="https://www.linkedin.com/in/shivdevrao/">https://www.linkedin.com/in/shivdevrao/</a></p><p>(3) <a href="https://www.bvp.com/atlas">https://www.bvp.com/atlas</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=120ac8b55e54" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What I read this week]]></title>
            <link>https://medium.com/@annamalaipushpavalli/what-i-read-this-week-be024b128dbd?source=rss-ed0c032a748c------2</link>
            <guid isPermaLink="false">https://medium.com/p/be024b128dbd</guid>
            <dc:creator><![CDATA[Pushpavalli Annamalai]]></dc:creator>
            <pubDate>Sun, 28 May 2023 20:32:35 GMT</pubDate>
            <atom:updated>2023-05-28T20:32:35.513Z</atom:updated>
            <content:encoded><![CDATA[<ul><li><a href="https://incubees.com/amazon-led-20-m-funding-received-by-kids-fashion-brand-hopscotch/">Amazon led $20 M funding received by kids&#39; fashion brand Hopscotch - Incubees</a></li><li><a href="https://www.entrepreneur.com/en-in/news-and-trends/edtech-platform-ulipsu-raises-32-million-funding/452876">Edtech Platform Ulipsu Raises $3.2 Million Funding | Entrepreneur</a></li><li><a href="https://www.usatoday.com/story/news/politics/2023/05/28/debt-ceiling-deal-details-biden-mccarthy-takeaways/70264836007/">What&#39;s in Biden and McCarthy&#39;s debt ceiling deal − and what&#39;s not? Here&#39;s 7 takeaways</a></li><li><a href="https://www.wsj.com/articles/the-great-strategist-henry-kissinger-turns-100-china-ukraine-realpolitik-81b6f3bb?st=ezd7s7y8rkmqrm9&amp;reflink=article_copyURL_share">Opinion | Henry Kissinger Surveys the World as He Turns 100</a></li><li><a href="https://www.wsj.com/articles/how-ai-is-catapulting-nvidia-toward-the-1-trillion-club-14f42380">How AI Is Catapulting Nvidia Toward the $1 Trillion Club</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=be024b128dbd" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>