Operations Owns the Substrate. Most of Them Don't Know That Yet.

Picture the standing meeting where AI strategy gets discussed in your business. Who’s in the room?

Probably the CTO or Head of IT. Probably a few of their team. Maybe the CEO. Maybe a vendor running a demo. Almost certainly somebody with a slide deck called “AI Roadmap” or similar.

Now picture who isn’t in the room.

The head of operations. The customer service director. The training lead. The compliance team. The quality manager. The people who, between them, own every word of the content your AI will actually read from.

This is the gap. It’s everywhere and almost no one is talking about it.

I currently work in operations and quality leadership in regulated financial services. I’ve watched this gap play out across teams. IT is buying the AI runtime. Operations is writing the content that powers it. Neither side knows the other is half the equation.

Why the gap exists

AI looks like a technology project. So it’s defaulted to IT, the way every technology project does.

If you ask a Head of IT what their AI strategy is, they’ll talk about agents, MCP servers, vector databases, vendors they’re evaluating. They’ll talk about retrieval-augmented generation. They’ll talk about which large language model they’re using, or which agent orchestrator. They are not making this up. These are real, important decisions.

If you ask a Head of Operations what their AI strategy is, you’ll often get a polite shrug. “That’s an IT thing.” Or “we’ll see what they roll out.” Or “we’re getting some training pushed through later this quarter.”

This is a category error. Both sides have been told it’s the other team’s problem. And it isn’t.

What Operations actually owns

Walk through what Operations produces in any given week.

Standard operating procedures. Quality manuals. Training materials. Compliance content. Customer-facing scripts. Escalation matrices. Policy updates. The answer to “what do we do when X happens?” across every team you run.

This is the substrate. This is what the AI runs on.

Every AI agent, every chatbot, every customer-facing assistant, every internal copilot will produce answers based on what your Operations team has written or hasn’t. If Operations content is current, structured, owned by named humans, the AI does well. If it’s stale, scattered, conflicting, or simply missing, the AI does badly.

You don’t change that by buying a smarter AI. You change it by treating Operations content as the substrate, and Operations as the team that owns it.

What goes wrong when Operations isn’t in the room

You know what happens next.

IT buys an AI tool. Plugs it into whatever content the company has. Notices the answers are bad. Hallucinations. Wrong fee structures. Outdated policy quotations. Confidently delivered guesses.

IT goes back to the vendor and asks for a smarter AI. The vendor sells them another model, or another retrieval setup, or another bolt-on. The quality improves a bit, then stalls, because the actual ceiling is the content.

The content is Operations’ domain. The conversation about how to make the AI better is the wrong conversation, in the wrong room, with the wrong people.

MCP, in plain English

Here’s the bit that makes the Operations leader’s job a lot easier.

MCP stands for Model Context Protocol. Pretend it doesn’t. The technical name is unhelpful for what it actually does.

What MCP actually is: it’s the wire that lets your AI tools read from a canonical source instead of inferring from whatever content is lying around. Think of it like a power outlet. Your kettle doesn’t care which power station the electricity came from. Your AI doesn’t need to be re-engineered for each new content source. The interface is standard. Plug in. Power on.

If your Operations content lives in a substrate, any AI tool that speaks MCP can plug into that substrate and get the answer. Same content, every tool, every agent, no rebuild.

Why this matters for Operations: the work you do to clean up your content has compounding value. Update one article, every AI tool that reads from the substrate gets the new version on its next call. No IT project. No retraining. No “we’ll roll it out next quarter.” The substrate is yours. The runtime is whatever IT is using this month. Both can change, and the work doesn’t get redone.

If you want to see what that actually looks like, send me a message. I’d rather walk you through a quick demo of how KnowledgeScout works for an Operations leader than explain it any further on the page. It’s faster to see it than to read about it.

The case for being in the room

Once you’ve seen the picture, the decisions IT is making about AI are content decisions in disguise.

Where does the AI read from? That’s a question about what counts as the source of truth in your business.

Who owns the answer? That’s a question about who has accountability for what the AI tells your customers and your staff.

When does an answer get updated? That’s a question about your content lifecycle and change management.

How do we know the answer is current? That’s a question about freshness signals, review dates, ownership.

These are Operations questions. They have always been Operations questions. They’ve been getting answered by IT mostly because IT is the only team in the room when the questions get asked.

The honest admission

This isn’t a turf war. IT teams aren’t villains. They’re solving the half of the problem they can see.

Operations teams have been so busy producing content that they haven’t looked up to see what IT is doing with it. Both sides have an honest reason for the gap. Neither side has been wrong, exactly. They’ve just been working in separate rooms.

The fix isn’t to take work away from IT. It’s to put Operations in the meeting. Different perspective. Different questions. Better outcomes.

Where this is going

The next year of AI in your business is going to be loud. Vendors will pitch. The technology will keep changing. Some of the tools your business buys this year will be obsolete before the next budget cycle.

The substrate is the part that doesn’t change. The procedures, the policies, the scripts, the training content. Owned by humans with judgement. Reviewed on a cycle. Maintained as the source of truth.

The runtime will change. The substrate stays. Operations owns the substrate.

That’s the conversation Operations should be having. Not “when will IT roll out the AI?” but “is our substrate ready for whatever AI calls into it?”

Send me a message if you want to see what that looks like from the inside. The technology is easier than the marketing makes it sound. The hard part is who owns it. And that, finally, is your bit.

The KnowledgeScout Team