The board wants an AI strategy. Your team has ideas. A few vendors have already called. And somewhere between the pressure to act and the uncertainty about where to start, most companies do the same thing: they hire a consultant to write a deck.
Six months later, the deck is sitting in a shared drive. Nothing is in production. The original problem hasn’t moved. And now there’s a new round of pressure to show progress.
I’ve seen this enough times that I stopped being surprised by it. The gap between “we have an AI strategy” and “AI is actually working in our operations” isn’t a strategy gap. It’s a discovery gap.
The demo was real. The workflow never showed up.
There’s a specific moment that keeps happening in companies trying to move on AI. A vendor runs a demo. The tool looks impressive. The use case maps reasonably well to the problem at hand. Leadership gives the go-ahead to pilot.
Then it hits the security review. Then integration. Then the question of who actually owns it on the internal side. Six weeks later, the pilot is stalled, nobody is quite sure why, and the team goes back to the old way of doing things.
It’s not that the technology failed. It’s that nobody sat down with the people running the actual work, before the pilot started, and figured out whether this was the right problem to solve with AI in the first place.
Most assessments and strategies are written too far from the work. They’re built on org charts and leadership interviews, not on what happens between systems at the point where someone is copying data from a CRM into a spreadsheet at 4pm on a Thursday.
That’s the work that’s eating your best people’s time. And it’s almost invisible on a leadership dashboard.
What we kept getting asked to skip
When we started running AI Discovery engagements, clients occasionally pushed back on the structured approach. Five sessions felt like a lot when there was pressure to “just start building.” Senior pairing felt like overhead. Process mapping felt slow.
Every time we skipped steps to move faster, we ended up rebuilding. Not because the first build was technically wrong, but because we’d built the wrong thing with confidence.
The structured approach isn’t caution for caution’s sake. It’s what gets you to a roadmap you can actually defend, with real numbers behind it, built on what your systems and data will actually support.
The five sessions exist because the answer to “where does AI fit in our business” can’t come from a conference room. It has to be discovered inside the work, with the people who run it, by someone who can tell in the real world which opportunities will actually ship and which ones will stall at the integration review.
What Discovery actually produces
After two to three weeks, a senior pair embedded with your team, and five working sessions, what you have is not a deck. It’s a phased roadmap with three to six opportunities scored on value, feasibility, and data readiness, each with a T-shirt estimate so you can compare them honestly. A Phase 1 sized to deliver something real in weeks rather than quarters. And process documentation that stays useful long after the engagement ends.
The reason we score opportunities on data readiness alongside value is that some of the highest-value AI applications break down because the underlying data is inconsistent, siloed, or not captured at all. Finding that out before you’ve committed to a build is worth a lot. Finding it out after costs significantly more than the discovery did.
We also size Phase 1 specifically so it can move at board speed. Leadership needs to see something concrete, not just a plan. A well-sized Phase 1 gives you that, with a second phase ready to follow when it lands.
When we tell clients AI is the wrong answer
Fix the process first
We’ve turned down engagements where the real problem was a process that needed fixing before any automation made sense.
Plain automation over complexity
Sometimes the right answer is workflow automation with no AI involved, because AI would have added complexity without adding value.
Long-term relationships over revenue
Saying no is not how you maximize engagement revenue. It is how you build a relationship with a leadership team that will call you back when the right problem shows up.
Operations first, platforms second
Discovery starts with your actual operations, real data, existing systems, and the problems your people are trying to solve — then builds forward from there.
I’d rather be the team that told you not to build something than the team that built the wrong thing well.
If you’re being asked for an AI strategy
The pressure is real, and the question is legitimate. Your board is right that AI matters. Your competitors are moving. The risk of waiting too long is real.
What’s also real is the risk of moving on the wrong thing, in the wrong order, without the groundwork that makes it stick.
The most useful thing I can offer, before any conversation about what to build, is a clear-eyed answer to where AI will actually create value in your specific operation. Sometimes that answer points to Discovery. Sometimes it points somewhere else entirely. Either way, you leave with a clearer picture than you came in with, and that’s usually what people need most at this stage.