It usually arrives in a hallway or a Slack thread, phrased like it's obvious. "We need an AI plan." Sometimes it's the CEO after a board meeting; sometimes it's an investor update that name-checks a competitor. Either way the sentence lands on someone — often you — and it comes with no scope, no budget line, and a quiet assumption that the plan already exists somewhere and just needs writing down.
So you do the natural thing. You open a doc, you think hard about where AI could help, and you start a list from the top: customer support automation, a sales-email drafter, maybe something for finance. It feels productive. It is also, almost always, the wrong list — not because the ideas are bad, but because of where they came from. They came from the person furthest from the actual work.
Why the view from the top is the wrong altitude
The pressure is real, and it isn't imaginary hype. McKinsey's 2025 global survey found that a large majority of organizations now report using AI in at least one business function, and that the share using generative AI in particular has climbed steeply year over year — this is genuinely happening, not a rumor you can wait out. The problem is not whether to act. It's that the mandate creates urgency at exactly the altitude with the least detail.
From the top of an org, work looks like a neat set of boxes: Support, Sales, Finance, Ops. But the places AI actually saves time are not boxes — they're the seams inside them. The forty minutes someone spends every Monday reconciling two exports that don't agree. The copy-paste ritual between a form and a CRM that no process doc has ever mentioned. The report that one person rebuilds by hand each month because the template broke in 2023 and nobody had time to fix it. None of that is visible from the top. All of it is visible to the person doing it.
An AI plan written from the top is a list of what leadership imagines the work to be. The work itself is held, in vivid and unglamorous detail, in the heads of the people who do it every day.
The knowledge is already in the building
Here's the reframe worth sitting with. The most valuable input to your AI plan isn't a consultant's framework or a vendor's demo. It's the tacit knowledge of the people closest to each workflow — and there's a long research tradition on why that knowledge is so hard to extract from the top down. Management scholars have written for decades about tacit knowledge: the understanding people carry that they can act on fluently but struggle to articulate in a meeting, precisely because it lives in the doing. Ask a top-down question and you get top-down answers. The Monday-reconciliation problem only surfaces when you ask the person who lost the Monday.
There's a second, quieter reason the top-down list fails: it skips the people whose buy-in you'll need to make anything stick. When a plan is handed down, the people expected to change how they work had no hand in shaping it — and change research is unambiguous that participation is one of the strongest predictors of whether a change actually holds. A plan nobody helped build is a plan nobody defends.
What "aware" actually looks like
So the first honest move, before any tool or roadmap, is to change who you're asking. Not "leadership, what should our AI priorities be?" but "everyone who touches the work, what would you hand to a machine tomorrow if you could?" That single shift — from a plan authored at the top to one surfaced from the floor — is the difference between an AI initiative that produces a slide and one that produces saved hours.
You don't need to have the method figured out yet. You just need to notice the trap you're standing in: the mandate is real, the urgency is real, and the instinct to answer it alone from a doc is exactly what will produce a plan that looks complete and delivers nothing. The people who know the answer are already on your team. Over the next few posts we'll get into why the usual ways of collecting their input — the all-hands brainstorm, the suggestion box, the loudest-voice-wins meeting — quietly fail too, and what it takes to do it right.
For now, the shift is enough: stop writing the list. Start planning how to ask.