The Pilot Trap Is a Positioning Problem
McKinsey published a piece this summer, “Escaping the pilot trap” (opens in a new tab), about HR functions in the agentic era. The pattern it describes will be familiar to anyone who has watched a company adopt any new technology at all: a dozen pilots, each one individually defensible, each one fast, and no two of them building toward the same thing. Every one of them reports success, which is why it takes so long to notice that there is little additive progress.
I say: don’t scale the pilots, define the North Star for the function first. Then work backward from it to what you actually build. Settle the destination and the individual initiatives start adding up instead of merely accumulating.
I know, sounds so obvious, right?
The same advice is coming from BCG, Bain and Accenture about their own domains. BCG has a section heading for it: “From Scattered Pilots to a Shared North Star” (opens in a new tab). But I want to make a slightly larger claim about it than the reports do.
The pilot trap is not about AI
Strip the technology out of the description and look at what’s left. A company is running many small initiatives. Each one is justified on its own terms. But collectively they don’t add up, because the people running them are working from different assumptions about what the company is for.
That is not an AI failure mode. That is the ordinary condition of a company without a settled answer to what business it’s in. AI didn’t create this issue. AI just made it cheap enough to run twelve initiatives at once, so incoherence that used to take three years to become visible now takes two quarters.
Which is also why the fix is a North Star and not a governance framework. Governance catches the symptom. It will tell you that two teams built overlapping things. It won’t tell you which of them was right, because answering that requires knowing what you are trying to become.
The tell you can check this week
We’ve written before about asking ten people in your company to finish one sentence, separately and in writing: we help ___ do ___. Leaders predict minor variation in wording. What usually comes back is three or four genuinely different companies.
That test predicts the pilot trap. If those ten sentences describe four companies, then every initiative anyone launches is being launched against one of those four. The pilots aren’t disconnected because nobody coordinated them. They’re disconnected because they are solving for different businesses, and no amount of coordination fixes a disagreement the participants don’t know they’re having.
It’s the same reason the prioritization meeting never settles anything. People argue about sequencing when they are actually disagreeing about identity. Sequencing arguments are unwinnable when that’s the real subject, and everyone leaves the room feeling like the process failed.
What “define the North Star” actually asks of you
Here is where the advice in those reports tends to stop: everyone agrees the North Star comes first. But neither of those big consultancies say what a North Star is, who writes it, or what makes one good enough to work backward from.
The working test: a North Star has to be specific enough to rule things out. If it doesn’t kill a project somebody wanted, it isn’t doing any work. “Become an AI-first organization” rules nothing out, which is why it survives every meeting it enters. “We are the company that gets this buyer this outcome without this cost” rules out most of the roadmap. That is the point of it, and it’s also why it’s uncomfortable to write and easy to postpone.
It has to be written down, too, in language specific enough that someone can act on it without a meeting. A destination that lives in the founder’s head governs only the initiatives the founder personally reviews. That written form is the North Star document, and it is the artifact almost every engagement we run produces.
The order matters more than the speed
- Settle what business you’re in, in writing, specific enough to be wrong.
- Work backward from that to what you build.
- Then go fast.
Most companies start at three, hit the trap, and reach for governance. Which is step one arriving late, disguised as process, and considerably more expensive than it needed to be.
If your initiatives are all individually defensible and collectively going nowhere, that’s the conversation we have.