Your Board Wants an AI Strategy. Give Them an AI Budget Instead.

Your board asked for an AI strategy. You gave them 14 slides, and 6 months later not one of them has been held against you.

The ask usually lands near the end of a board meeting. An investor read something on the flight in, and it comes out as a polite question: "So what's our AI strategy?" The founder goes home and builds a deck. A vision slide, a map of 9 possible use cases, a grid of vendor logos, a roadmap with Q1 to Q4 swim lanes. It looks thorough. It commits to nothing.

The typical deck has a slide called "Our AI Opportunity." 9 boxes, each a use case, each with an arrow pointing at a bigger number. No owner on any of them, and no date. The board says "great progress" and asks the same question 2 meetings later.

The pressure is real, and it's rising. In a Harris Poll survey of 900 CEOs for Dataiku, 72% of US CEOs said their boards are pressing them to deliver measurable AI results, up from 61% a year earlier. Those are companies over $500 million in revenue, so your seed-stage board is a smaller animal. But the investors on your board sit on those boards too. They carry the question from room to room.

What a deck can't do

A strategy deck lists things you might do. Every item is free to write and impossible to fail, because none of them has a number or a date attached.

And the results look exactly like that. McKinsey's 2026 State of AI survey found that nearly 9 in 10 organizations use AI regularly in at least one function, but only 37% attribute any EBIT impact to it, about the same share as the year before. MIT's NANDA initiative was blunter: about 5% of AI pilots achieve rapid revenue acceleration, and the vast majority stall. That study rests on 150 interviews and 300 public deployments, so treat the precision as a headline. The direction matches what I see when founders call me.

Meanwhile the spending keeps climbing. Menlo Ventures estimates enterprises spent $37 billion on generative AI in 2025, up from $11.5 billion in 2024. Bain found that 42% of CFOs expect to raise AI budgets by 30% or more over the next 2 years, while only 31% are satisfied with their AI outcomes.

More money, same fog about what it bought. Boards see that gap. They say "strategy" because they don't have a better word for "show me this is under control."

So show them control. A budget is a decision your board can hold you to.

The one-page AI budget

Keep it to 1 page. If it spills onto a second, you're writing a deck again.

A fixed amount. One number for the period. Include tools, API spend, contractor time and the internal hours you're pulling off the roadmap. The hours are the line founders leave out, and they're usually the biggest.

1 to 3 named workflows. "Support ticket triage." "First-pass contract review." A named workflow has an owner and a before-state you can measure. "Use AI in sales" is a theme, and themes can't be audited. If you can't pick 3, pick 1. I wrote about how to choose that first workflow already.

One metric per workflow. Hours per week, cost per ticket, days to close. Write today's baseline next to it, because you won't be able to reconstruct it in 6 months.

A date. When the metric should have moved, and by how much. 90 days is a decent default. Long enough to get past the demo, short enough that the board still remembers what you promised.

Kill criteria, written in advance. "If cost per ticket hasn't dropped 20% by March 31, we stop and move the money." Write it now, while you're calm and before anyone has sunk cost to defend.

An owner. One name per workflow. A committee can't be held to anything.

That's 6 items. The kill criteria do most of the work.

How to report it every meeting

The budget only works if it shows up at every board meeting in the same format. Same page, updated. Directors should be able to compare this quarter to last quarter in 30 seconds.

For each workflow, 4 lines: spent to date against budget, the metric now against the baseline, a status (on track, at risk or killed) and what happens before the next meeting.

Killed has to be a normal status. If a workflow misses the criteria you wrote 90 days earlier and you shut it down on schedule, the budget did its job. Say so plainly, and show where the money went next. In my experience, a board trusts a founder who kills things on time far more than one who reports 9 initiatives "in progress" for a year.

Picture the update: "We spent $18,000 on automated invoice matching. The error rate never got under 4%, which was our line. We stopped it on schedule and moved the remaining $12,000 to support triage." That's a 30-second agenda item. The questions that follow get sharper, and the vague "AI strategy" question stops coming back.

For the workflows that survive, add one more line: who owns the judgment. I run the 70/30 split: agents do the mechanical 70%, and a senior human owns the 30% that needs judgment. Put that human's name on the page. It answers the risk question before a director has to ask it.

The hard part

Writing kill criteria feels like admitting, in writing and in front of investors, that you might fail. It is exactly that. It's also the one thing on the page they can't get from a vendor slide.

Founders who push back on this usually want to keep their options open. I get it. An open option with no number on it is also exactly what your board was worried about in the first place.

They asked for a strategy because they wanted proof you'd spend carefully. Hand them a number, a date and the conditions under which you'll stop, and you've given them one.