We had a call last week with a fashion brand that had spent £18,000 on an AI strategy engagement.
They got back a 64-slide deck, a maturity matrix, a workshop summary, an "implementation roadmap", and a recommendation to engage the same firm for "phase two: pilot selection". They had not shipped a single automation. They had paid the price of a service-tier build for a document.
This is not a one-off. We hear a version of it every month. So here is the operator-side accounting that nobody on the consulting side will write down.
The £18,000 was not the real cost
The cheque is the visible cost. The invisible costs are bigger.
Six weeks of leadership attention. Discovery interviews, workshop preparation, deck reviews, follow-up sessions, internal comms about the "AI initiative". Multiply two senior people by six weeks at fifty per cent attention and you have lost three person-months of operating capacity to a deliverable nobody outside the boardroom will read.
The cooling effect on the team. Six weeks of "we are working on AI strategy" tells everyone in the building that AI is something that happens at boardroom altitude. By the time you are ready to actually do something, the people who would have run with it are mentally checked out.
The phase-two trap. A deck does not solve a problem. So you need phase two. Pilots. Proofs of concept. Vendor selection. Each one is more invoices, more weeks, more attention. The deck is not the deliverable. The deck is the gateway drug for an annuity.
The opportunity cost of a working system. Six weeks and £18,000 buys you, with a competent operator, two services live in production. Cart recovery and lead qualification, say. Both running. Both saving hours. Both proving the thesis with their own results.
That is the real cost: you traded compounding outcomes for a static document.
What strategy decks are good at
To be fair, strategy decks do solve a problem. Just not the one you think.
They solve "I am responsible for AI internally and I need cover for the budget I am about to spend." They produce committee-ready materials. They de-risk the political case for the next round of consulting. They give the procurement team a framework to evaluate vendors against.
If you are a Fortune 500 with internal politics to navigate, that has real value. If you are a small business operator trying to recover £8k a month in abandoned carts, you are buying expensive cover for a problem you do not have.
A deck does not solve a problem. So you need phase two. The deck is not the deliverable. The deck is the gateway drug for an annuity.
What to demand instead
If you have a working budget for AI and you want to know whether to spend it, here are the questions to ask any vendor before you sign anything.
"What will be live by the end of week two?" If the answer is anything other than a specific automation running against your real data, you are buying a discovery engagement, not a system. Walk.
"Show me a working system you built for someone like me." Not a case study deck. A demo, ideally screen-shared, of the thing actually running. If they cannot show you working software, they do not build working software.
"What will you ship with me, and what will you keep?" Strategy firms keep the methodology and rent it back. Operators ship you the workflow and walk away. Your stack should not require their continued involvement to function.
"What is the success metric, and when do we measure it?" Real builds have measurable outcomes attached. £6,000 a month in recovered carts. 70% of support queries auto-resolved. Two hours per day of admin returned. Without numbers, you are buying activity, not outcomes.
The cheaper, better alternative
There is a thing called "doing it" that almost no consultancy will sell you, because doing it costs them margin and creates the wrong precedent for the next client.
We sell doing it. So do a few other operator-led shops. We charge less than half what a strategy engagement costs, and at the end of two weeks you have a system running. We are not braver or smarter for this. We just made a different bet about what marketing-services-economics looks like in the era of capable AI.
If you are about to commission a deck, do this first. Pick the one process in your business that is currently bleeding the most time or money. Just one. Brief us, or anyone like us, on that single problem. Spend a quarter of the deck budget. Two weeks later, see whether it works.
If it does, you have your strategy. The strategy is "do that, eight more times, in priority order." Total deck count required: zero.
If it does not, you have lost a quarter of the budget and you have learnt something specific about why automation does not fit this particular problem. That information is worth having.
Either way, you are ahead of the team that is on slide 47.