“Tom, we are using AI, but we are not seeing the right results. We sound like everyone else.”

I hear this more than almost anything right now. And the founder saying it is not wrong. Their proposals feel interchangeable. Their content is polished, but hollow. Their sales process is faster, and somehow less effective. They have invested in the tools. They have trained the team. They are generating more output than ever.

None of it is landing. And it has nothing to do with AI slop.

It is AI bilge.

Bilge is the lowest part of a ship’s hull, the curved underbelly where everything collects. The water that pools there is bilgewater: stagnant, contaminated, waste from everything above.

But bilge has another meaning entirely: nonsense; rubbish. Output that sounds plausible and means nothing. Many businesses are producing both with AI.

The real problem is not the tools

Every business in your market has access to the same models. The same platforms. The same capability. The differentiator is not which tools you use. It is what those tools have to work with.

AI does not generate commercial intelligence from thin air. It reflects back what it is given.

Feed it a business with clearly defined positioning, documented processes, explicit decision frameworks, and it produces outputs that sound like you, sell like you, and compound over time. Feed it a business held together by founder gut instinct, and it produces impressive-looking outputs that could have come from anyone.

Here is what that looks like in practice:

Without architecture

A founder opens an AI tool and types: “Write a proposal introduction for a professional services firm.” They get something generic, adjust the tone, add client context, rewrite the positioning section entirely, and spend forty minutes on something that was supposed to take five. The AI did not fail. It had nothing real to work with.

With architecture

Another founder does the same task, but pastes in a documented positioning statement, a defined ideal client profile, and a clear articulation of why clients choose them over alternatives. The output is usable in minutes. Not because the model is different. Because the architecture underneath it is.

Most founder-led businesses are firmly in the first category. Not because they are poorly run. Because they have grown by doing, not by documenting. The founder carries the commercial logic in their head. It works, until they try to scale it.

HBR published research on this earlier this year: “When Every Company Can Use the Same AI Models, Context Becomes a Competitive Advantage.” The title is almost the argument. The differentiator is not access. It is what is already built underneath. AI does not solve that problem. It inherits it.

What commercial architecture actually means

Commercial architecture is the infrastructure underneath everything: how decisions get made and by whom, how value is defined and communicated, how the commercial process runs when the founder is not in the room.

It is the difference between a business that has a sales process and one where the founder is the sales process. Between a business that can articulate why clients choose them and one where “it is relationships” is the best answer anyone can give.

I grew a compliance and training consultancy over fifteen years before it was acquired. One of the clearest signals that the commercial architecture was working came not from a sales conversation, but from an inbound call. A leading insurer approached us about a major tender, not because we had pitched them, but because we had spent two years being the only firm consistently visible at the right industry events, with a differentiated proposition they could actually point to, and systems that looked like they could handle scale.

They did not choose us because we were good at compliance. Plenty of firms were good at compliance. They chose us because we looked like a business that could handle the opportunity. That is commercial architecture doing its job. It does not just make your AI outputs better. It makes you the obvious choice before a conversation even starts.

If you are constantly correcting AI outputs, rewriting the positioning section, adding client context, changing the tone to sound less generic, that is not an AI problem. That is what it looks like when your commercial logic lives exclusively in your head.

The AI is just making that visible. It was always the problem. It is just more expensive now.

What separates the businesses pulling away

The businesses generating genuine commercial advantage from AI right now share a few characteristics that have nothing to do with the tools they have chosen.

They know exactly who they are for and why those clients choose them. Not in aspirational terms, but in specific, uncomfortable, provable terms. They have documented their commercial process thoroughly enough that an AI prompt can be given genuine context. They have defined decision rights clearly enough that outputs can be reviewed and deployed without the founder in the loop for every piece.

They built the architecture first. AI runs on top of it.

The result is not just better AI outputs. It is a business that can operate, communicate, and sell consistently, whether the founder is present or not. AI accelerates that consistency. It does not create it.

The question worth sitting with

If every AI output your business produces requires significant human intervention to sound right, ask why.

Not, “why is the AI getting it wrong?” Ask why a tool that has access to everything in your business still cannot speak for it. Ask what that tells you about what has actually been built. Ask whether the businesses pulling away from you in the next 18 months are asking the same question, or whether they have already answered it.

The founders who understand that distinction are already moving. Probably faster than you. That should be your real concern.