The pattern is so consistent it’s almost a tell
A mid-market company decides it needs an AI strategy. They engage a recognizable consultancy. Twelve weeks later they get a 60-page deck, a beautiful PowerPoint roadmap, an opportunity matrix that scored 47 use cases on a 2×2, and an invoice for somewhere between $250k and $1.2M.
Then comes the separate statement of work. The implementation engagement. Two to four times the original fee. Subcontracted — quietly — to a body shop you’ve never heard of, with engineers who weren’t in the original strategy room.
Six months later, the strategy partner has moved on to the next account. The body shop has shipped a half-finished pilot. Nothing is in production. The executive who sponsored the engagement is updating their resume.
This pattern is so consistent across the industry that it’s not bad luck. It’s the business model.
Why the structure forces this outcome
Consultancies optimize for billable hours per partner. A partner billing at $1,200/hour cannot economically write code — the opportunity cost is too high. So the partner sells strategy. The strategy work generates the slide deck. The slide deck is the deliverable. The implementation gets pushed to the next contract, where the partner can stay billable on “executive oversight” while juniors do the work.
This isn’t a moral failure. It’s incentive design. If you priced senior partner time at $200/hour you’d get partners writing code. You’d also get a consultancy that goes out of business. The economics demand slides.
The problem is that AI doesn’t behave like an ERP rollout. With an ERP, the strategy and the implementation can be cleanly separated — the vendor is named, the modules are known, the configuration is parameterizable. With AI, the strategy and the implementation are the same conversation. You don’t know if a use case is feasible until you’ve tried it on real data with real users. The deck is wrong by month three because the model landscape moved. The capability gap analysis is wrong because nobody on the strategy team has actually built a multi-agent system in production.
What to demand from your next engagement
A few simple tests will tell you whether you’re hiring strategy theater or actual shipping capability:
- Will the same senior person who writes the roadmap also write code in your environment? If the answer is no, you’re hiring strategy theater.
- What’s the deliverable at week 8? “A roadmap document” is the wrong answer. “A running system” is the right one.
- Show me a system you built that’s running in production today. Not for a client (NDA). For yourselves. Your own infrastructure, your own dogfooding. If the consultancy doesn’t run its own AI systems, they don’t know what production AI feels like.
- What’s the engagement structure if month 3 isn’t going well? A team that ships code can pivot. A team that ships slides will defend the slides.
- How do you decide between Claude, GPT, and open-source models? If they have one answer, they have a vendor relationship. The honest answer is “depends on the use case, here’s the matrix.”
The fractional model is the structural fix
The reason fractional AI leadership works is that one senior person owns both the strategy and the execution. There’s nowhere to hand off to. The roadmap they write is the roadmap they ship against. The architecture they recommend is the architecture they commit. The vendor pitches they vet are the contracts they review.
You can’t hide behind a body shop because there is no body shop. You can’t push the implementation to a separate SOW because you’re already implementing. You can’t ship a slide deck because the deliverable is a running system.
It’s also dramatically cheaper. A senior fractional AI leader at 2 days a week costs a fraction of a Big 4 engagement and produces working code. The trade-off is that you have to actually let them into your business — sit them in your ops meetings, give them production access, treat them like a senior member of the team. That’s uncomfortable for some executives. It’s also the only way real AI work gets done.
The hard truth
The companies that will look back on 2025–2027 as the period where they built durable AI advantage are the ones that figured out, early, that strategy without execution is procrastination. They hired senior builders. They ran short engagements with hard milestones. They demanded code in production, not roadmaps in PowerPoint.
The companies that will spend the same period burning consulting dollars and showing nothing for it are the ones still in the strategy-deck loop.
You probably know which category your last engagement fell into.
