Why 'AI strategy' decks ship nothing
Every consulting firm in the world is selling AI strategy. Almost none of it ships. Here's why, and what to demand instead.
There's a version of AI consulting that's eating budgets alive right now. It looks like this: a six-figure engagement, a multi-month assessment, a beautifully designed strategy deck, a 'roadmap,' a 'governance framework,' and a steering committee. At the end, nothing ships. The team is more confused than when they started, and the budget is gone.
This is not a niche pattern. It's the default mode of how legacy consulting firms have rebranded their old transformation playbook with the word AI on top. Same deck, new vocabulary, same lack of working software at the end.
Why the deck doesn't ship
Three reasons, structural:
- The team writing the deck is not the team that would build the software — different skill sets, different incentives
- The deck's job is to be approved, not to be built — the language is generic enough to survive a board review, which makes it useless to an engineering team
- The 'roadmap' lists 30+ initiatives — anything with that many priorities has none
When the consulting firm hands off, the in-house team looks at 80 slides and asks 'what do we ship Monday?' The deck doesn't answer. It was never meant to.
What to demand instead
Replace the deck-and-roadmap engagement with a ship-something engagement. Specifically:
- A working agent or workflow in production at the end of week 4, not a roadmap
- Code, prompts, and integrations the in-house team owns and can edit
- A defined success metric agreed to before the engagement starts
- A human in the loop on every workflow that touches a customer until the team trusts the output
- Knowledge transfer baked into the build process — pair, don't hand off
This is the inverse of the deck-first approach. It's also exactly the model behind the five AI workflows we ship first.
The 'we need to align first' trap
The objection from deck-first consultants is always the same: 'you can't just build, you need alignment first.' This is true at $50B enterprise scale. It is almost never true at $5–100M service-business scale, where the founder is the steering committee.
Service-business operators don't need a 12-week alignment exercise. They need to see a working agent score a real lead and write a real summary. Once that exists, alignment happens in a 30-minute meeting because the thing is real and obviously valuable.
The fastest way to align a company on AI is to ship one working agent. The slowest way is to write a 60-slide deck about why you should.
How to evaluate an AI partner
- Ask to see something they shipped in the last 90 days — code, agent, workflow, anything live
- Ask what their first 30 days look like — if there's no working software at the end of it, walk
- Ask who owns the code and prompts at the end — must be you
- Ask how knowledge transfers — pair programming and shared repos, not slide decks
- Ask their position on human-in-the-loop — anyone proposing fully autonomous agents on day one is wrong
What a real engagement looks like
- Week 1: Pick one workflow. Define the success metric. Map the existing process.
- Week 2: Build the agent or automation. Get it working on test data.
- Week 3: Deploy with a human approval step. Run on real volume.
- Week 4: Measure against the success metric. Decide whether to expand or iterate.
- Months 2–6: Repeat with the next workflow. Build the team's muscle.
This is the model behind our AI integration work. The deliverable is software, not slides. The clients in our case studies stayed with us because the agents kept shipping, not because the strategy decks were polished.
The bottom line
If your AI engagement ends with a deck instead of a working agent, you didn't buy AI — you bought consulting theatre with the word AI on the cover. Demand software. Demand ownership. Demand a 30-day timeline. Anyone who can't deliver against that doesn't actually build, they just present. And the five workflows piece is the right next read on what to ship first.