Two years ago, I wrote about the four types of enterprise AI entering the market. The thesis was straightforward: the AI categories that actually create competitive advantage — enhanced workflows and enterprise-connected AI — all depend on solving a problem most organizations still haven’t touched.
Integration.
I’ve been beating this drum for a while. In 2023, I wrote It’s Groundhog Day and warned that the same integration bottleneck that killed mobile and digital transformation initiatives would come back to haunt AI. Later that year, we published an entire white paper on what we called the “Organizational OS” — the argument that composable integration wasn’t just nice-to-have, it was the operating system that would determine whether AI could actually function inside a business. I quoted Ross Mason’s line about AI without connectivity being a “brain in a jar.”
That was almost three years ago.
The thesis has more than held up. But what’s happened since has been more dramatic than I expected.
The Models Are Not the Problem Anymore
In 2024, everyone was chasing models. Who had the best LLM? Which vendor’s copilot was smartest? That was the shiny object.
Two years later, the model wars are largely settled. GPT, Claude, Gemini — pick your favorite. They can reason, write code, analyze documents, hold multi-step conversations. Astonishingly capable, all of them.
And yet most organizations still aren’t getting transformative value from AI.
The reason isn’t the model. It’s the plumbing.
Every App Is Getting Smarter. None of Them Talk to Each Other.
Here’s what happened while everyone was shopping for AI: the typical SMB’s tech stack quietly became a minefield.
Depending on who’s counting, somewhere between 100 and 250 SaaS applications — and that’s before you count the ones your team is using without IT knowing about it. CRM, ERP, HRIS, marketing automation, support, accounting, and dozens more nobody’s tracking. Each with its own data model, its own API (if you’re lucky), its own version of the truth.
Now every major vendor is embedding AI into their product. Salesforce has Agentforce. Microsoft has Copilot. HubSpot, ServiceNow, Workday — everyone has an AI story now. And each one is intelligent within its own walls.
But none of them can see the whole business.
I’ve heard people in the ecosystem call this the “SaaSpocalypse” — not a collapse of SaaS, but an explosion of intelligent silos. Every application getting smarter in isolation. The gaps between them getting wider. Your AI tools can each see a piece of the picture, but no single system can see the whole thing.
For SMBs, this is an existential strategic problem disguised as a technical one.
Why SMBs Should Care Most
Enterprise companies have been wrestling with integration for decades. They have middleware teams and seven-figure budgets. They’ll adapt.
SMBs don’t have that luxury. And here’s the irony: SMBs have the most to gain from AI. AI is the thing that lets a 50-person company operate like a 500-person company. Automating workflows, eliminating manual data entry, giving every employee access to insights that used to require an analyst team.
But AI can only act on what it can see.
An AI agent that can’t access your order data can’t help with fulfillment. An AI assistant that can’t read your CRM and your ERP simultaneously can’t answer a real question about customer profitability. An AI-powered quoting tool that can’t pull from your product catalog in real time is just a chatbot with opinions.
The integration layer is not a prerequisite for AI. It is the AI strategy.
None of This Is New. The Stakes Are.
If you’ve followed my writing over the years, you know I’ve seen this pattern before. Mobile disrupted businesses in 2013 — and the biggest blocker wasn’t the UI framework. It was getting the right data to the right place at the right time. I lived it firsthand building the #1 mobile experience in retail at HHGregg.
Salesforce implementations? Same story. Just about every conversation I have with a customer who’s frustrated with their CRM comes down to lack of access to data in other systems. The platform is incredible. The build is what fails — and it fails because of integration.
AI is no different. Except this time, the companies that solve it don’t just get a better mobile app or a better CRM. They get a fundamentally different way of operating.
With integration: one connected environment where AI reasons across your entire business. An agent that can see a customer’s support history, contract terms, open invoices, and product usage — all at once — and take action. That’s not a productivity improvement. That’s a different company.
Without it: five different AIs, each smart about one thing, none of them able to coordinate. Your team is still the integration layer.
MuleSoft exists to solve this. It connects any system to any system through a managed, reusable API layer. In the AI era, what that really means is giving AI the ability to see and act across your entire business — not just within one application’s walls.
What to Do About It
Stop thinking about AI as a product you buy. The value comes from connecting AI to your specific business context. That requires integration, not license upgrades.
Map where your data actually lives. If the answer is “spread across eight systems that don’t talk to each other,” you’ve found your AI bottleneck.
Treat integration as strategic infrastructure. This isn’t an IT project. It determines whether your AI investments pay off or stall.
Move fast. Most of your competitors are still running pilots in isolation. Every month of production AI generates compounding advantages. Waiting to “see how AI plays out” is a strategy for falling behind.
We Live This
I’ll be transparent: this isn’t theoretical for us. Green Irony runs on this architecture — MuleSoft connecting AI to our business systems — and the result is a small team operating with the speed and scale of a company many times our size. We build the same thing for our customers.
More on that soon. For now, the takeaway: the road to AI-native doesn’t start with picking the right model. It starts with connecting your business.
Everything else follows.
Aaron Godby is the CEO of Green Irony, a Salesforce and MuleSoft consulting partner specializing in AI-led delivery for small and mid-size businesses.