The Consultancy Trap: Why AI Advisory Is Full of People Who've Never Built Anything

The first thing most AI consultancies produce is a maturity model. You've seen them: a five-stage framework that places your organisation somewhere between 'ad hoc' and 'optimised,' accompanied by a reassuring recommendation that you need to move from stage two to stage four, and here's the roadmap to get there.

These models aren't useless, but they're deeply insufficient. They describe states without explaining transitions. They benchmark against an idealised end state that may not be relevant to your specific context. And they create a false sense of progress: moving from 'emerging' to 'established' on a maturity model feels like achievement, even when nothing has actually changed in production.

The consultancies that produce these frameworks are selling certainty in an uncertain domain. The maturity model is comforting because it implies a predictable path. But AI transformation isn't predictable. It's messy, iterative, and full of surprises that no framework anticipated — because the people who wrote the framework never encountered those surprises themselves.

When your AI consultant has never built a production system, their advice is necessarily theoretical. They can tell you what good looks like based on case studies and frameworks. They can't tell you what goes wrong in practice, because they've never been there when it did.

This matters because the hardest part of AI isn't choosing the right model or designing the right architecture. It's the operational reality: data quality issues that only emerge at scale, integration challenges with legacy systems, user adoption problems that no change management framework predicted, and governance requirements that conflict with delivery timelines.

An adviser who's only ever worked at the strategy layer doesn't see these problems until they've already derailed the project. An adviser who's built production systems recognises the warning signs early — because they've been burned by them before.

I don't say this to be provocative for its own sake. I say it because I've watched organisations spend millions on AI strategies from firms that couldn't write a line of code, only to discover that the strategy was architecturally infeasible, operationally naive, or commercially disconnected from reality. The advice was intellectually sound. It just didn't survive contact with the real world.

Galahad exists because of this gap. We built it as the consultancy we wished existed when we were on the client side — one where the people advising on AI strategy are the same people building AI systems.

Our advisory work is grounded in our products: Grail for generative engine optimisation, Lancelot for media quality and fraud detection, Enable for AI capability building. These aren't theoretical offerings. They're production systems that we build, maintain, and operate. When we advise a client on GEO strategy, we've built the platform that measures it. When we advise on media transparency, we've built the system that detects fraud. When we design a training programme, we've built the platform that delivers it.

This isn't about selling products through consulting or consulting through products. It's about ensuring that every piece of advice we give has been tested against the reality of building and shipping. The feedback loop between building and advising is what makes the advice credible — because we discover the problems, the edge cases, and the failure modes in our own work before we encounter them in client engagements.

If you're evaluating AI consultancies, ask three questions. First: what have you built? Not what have you advised on — what have you actually shipped to production? If the answer is a list of client engagements with no owned technology, that tells you something about the depth of their technical understanding.

Second: who will be on my team? Not the partner who sold the engagement — the people who will actually do the work. What's their background? Have they built AI systems, or have they built decks about AI systems? The gap between these two profiles is the gap between advice that works and advice that sounds right.

Third: what went wrong on your last engagement? A consultancy that can't articulate failure hasn't been close enough to execution to encounter it. The best advisers are the ones who've learned from their own mistakes — in their own systems, not just in client post-mortems.

The AI consulting market is growing fast, and much of that growth is firms repackaging existing management consulting as 'AI strategy.' Caveat emptor. The organisations that choose advisers based on building credentials rather than brand prestige will get better outcomes — because the advice will be grounded in reality, not in frameworks.