<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom"><title>Ollie Fritz - thoughts</title><link href="https://olliefritz.com/" rel="alternate"/><link href="https://olliefritz.com/feeds/thoughts.atom.xml" rel="self"/><id>https://olliefritz.com/</id><updated>2026-04-22T00:00:00+01:00</updated><entry><title>Thoughts on the Future of AI in Enterprise: You Shouldn’t Just Be the Interface</title><link href="https://olliefritz.com/writing/ai-in-enterprise-dont-be-the-interface.html" rel="alternate"/><published>2026-04-22T00:00:00+01:00</published><updated>2026-04-22T00:00:00+01:00</updated><author><name>ollie</name></author><id>tag:olliefritz.com,2026-04-22:/writing/ai-in-enterprise-dont-be-the-interface.html</id><summary type="html">&lt;p&gt;As hyperscalers dominate enterprise AI, B2B SaaS companies should focus on embedding domain-specific AI tools and designing for seamless integration with enterprise AI stacks.&lt;/p&gt;</summary><content type="html">&lt;p&gt;This is not a novel take, but one that is becoming ever clearer to me. Unless you’re an “AI company” in the purest sense, the future isn’t about being the AI interface for enterprises. It’s about being an indispensable tool &lt;em&gt;within&lt;/em&gt; an enterprise’s broader AI ecosystem.&lt;/p&gt;
&lt;p&gt;This idea hit home again at the &lt;strong&gt;London AI Breakfast Club&lt;/strong&gt; focused on B2B SaaS impact. The CEO of Duco made a blunt point: &lt;em&gt;“Hyperscalers will be the only providers large enterprises choose in the long run.”&lt;/em&gt; That aligns with what I have been expecting - embedding a chatbot in your SaaS app to “do the work” isn’t a sustainable strategy.&lt;/p&gt;
&lt;p&gt;Instead, an B2B SaaS company that focuses on large enterprises should focus on two things:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;Own your domain expertise (and data). Build embedded AI into your product where you can be the arbiter of models and functionality. Think transaction classification, reconciliation, specialist prediction or forecasting, or any use case where your IP and deep understanding of the problem gives you an edge.&lt;/li&gt;
&lt;li&gt;Design for integration. Create the right architecture to let enterprises plug your system into their AI stack. Standards like MCP and A2A are becoming table stakes [1], just as API-first design did a decade ago. IT departments will soon mandate that all systems integrate seamlessly into their AI workflows [citation needed].&lt;blockquote&gt;
&lt;p&gt;[1] I know there is a whole MCP vs CLI vs A2A vs whatever debate raging right now. Not saying that these are the end-all-be-all tools, but it is what current providers want.&lt;/p&gt;
&lt;/blockquote&gt;
&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The age-old question remains (AI has not changed that): &lt;em&gt;What’s your unique selling point? What is your comparative advantage?&lt;/em&gt; Unless you’re building a better general-purpose agent than a multi-billion dollar AI company (and let’s be real - you’re probably not), your advantage lies in solving a specific problem &lt;em&gt;within your domain&lt;/em&gt; in a way that complements the giants.&lt;/p&gt;
&lt;p&gt;&lt;strong&gt;Bottom line:&lt;/strong&gt; The future isn’t about being the AI interface. It’s about being the tool that &lt;em&gt;enables&lt;/em&gt; AI to deliver real value. At least in my opinion.&lt;/p&gt;</content><category term="thoughts"/><category term="ai"/><category term="thoughts"/></entry></feed>