The Beliefs That Make AI Advice in London Expensive and Useless

• Myth: A strategy deck is the deliverable
• Myth: The platform vendor will give you a straight build-versus-buy answer
• Myth: A working demo means the feature is nearly done
• Myth: AI belongs on the roadmap because investors expect it
• What good looks like

A lot of money changes hands in London for AI advice that never survives contact with a build. The pattern is consistent enough to be worth naming: teams buy a document, feel reassured, and then discover months later that nobody can act on it. The myths below are the ones that keep that cycle going.

§Myth: A strategy deck is the deliverable

Workshops feel productive. Whiteboards fill up, everyone aligns on ambition, and six weeks later a slide deck lands. It describes where you want to be. It rarely describes how you get there, what it costs, or what has to be true first.

The gap shows up in the detail. A roadmap without architecture, a cost model, or a sequenced set of dependencies is not a plan — it is a wish with formatting. Engineering teams read these documents and quietly disagree with half of them, then go back to what they were doing. Nothing gets built, and the budget is gone.

A useful output looks different. It is something a development team could pick up and start on without a follow-up meeting to decode it. That is a much lower bar than it sounds, and far more engagements miss it than admit to it.

§Myth: The platform vendor will give you a straight build-versus-buy answer

Build versus buy is probably the most consequential decision in an AI project, and it is routinely handed to someone whose revenue depends on the outcome. Ask a platform which platform you need and the answer is never a surprise.

The problem is not that vendors lie. It is that incentives shape framing. A recommendation to wait, or to build a narrow piece in-house and leave the rest alone, does not appear in a sales deck. Neither does an honest account of where a bought solution stops fitting once your volume or accuracy requirements shift.

Good advice on this question comes with reasoning attached: what each option costs, where the answer flips, and what would have to change for the recommendation to reverse. If that reasoning is missing, the recommendation is not advice.

§Myth: A working demo means the feature is nearly done

The demo is the easy part. A prototype that answers questions in a review meeting proves the concept exists, not that it holds up. Accuracy often turns out worse than anyone measured, inference costs do not scale the way the business case assumed, and the feature ends up in a branch nobody wants to own.

This is where a second opinion earns its keep — a senior read on the architecture, the evaluation, and whether finishing is worth the spend. The awkward version of that conversation is more valuable than the comfortable one.

§Myth: AI belongs on the roadmap because investors expect it

Adding AI to a funding narrative is not the same as adding it for users. Teams that start from the investor slide tend to end up with a feature that exists to be mentioned rather than used, and the maintenance cost lands on engineers who never wanted it.

The question worth asking first is narrower: where does this genuinely fit, and where does it not? A consultant who will tell you that some of your product should stay conventional is more useful than one who agrees everything needs a model.

§What good looks like

An honest engagement ends with something concrete and defensible: a written answer your board will not pick apart, a clear view on whether your data and team are ready, and a recommendation you would act on yourself. It also comes with a straight account of who the adviser works well with — and who they don’t.

The full guide, including how those engagements are structured and what each one covers, is on AI consulting from engineers who ship.