How to Pitch an AI Pilot and Actually Get the Budget

Pitches missing any of three specific things tend to stall — and not because the idea itself is bad. This applies whether you’re the one pitching an AI pilot, or you’re the one who needs to evaluate someone else’s pitch and decide whether to fund it.

The first is a named owner — one person fully accountable for the spend and the outcome, whether that’s you, your boss, or someone else. This sounds obvious, but time and again, automation projects without a clearly accountable owner stall out. AI has made the first step deceptively easy — anyone can spin up a working demo now, get the dopamine hit of a cool prototype, get patted on the back for it. Turning that into something genuinely productionized, with the controls, governance, and process wrapped around it that make it usable by other people, is an entirely different project, and it needs someone whose job it actually is to see it through.

The second is a dollar ceiling. These efforts always carry an R&D component, and having an approved budget — even a small one — makes an enormous difference in how far you can actually get. I recently ran a go-to-market project involving enriching client data and asked for $200 to test it. Got approved immediately, didn’t have to think about it again, and that was enough to prove out whether an enrichment approach actually worked against a few different providers. Without that approved ceiling, I’d have been stuck not knowing whether I was allowed to spend anything at all — which is exactly the situation that quietly kills a lot of promising pilots before they get a real test.

The third is a checkpoint date. Not a fixed “done by” deadline — just a real, scheduled point to check in with whoever approved the spend and tell them how it’s going. Other priorities always creep in. Something might be working great and just need more budget. Something might have stalled because you haven’t had time to focus on it and need help reprioritizing. Either way, having that date already on the calendar means you’re not scrambling to find thirty minutes with a stakeholder two days before you need it.

Worth naming directly: AI spend doesn’t behave like traditional IT spend. It’s not a predictable, depreciating asset purchase the way a software seat is. It behaves more like funding a series of fast experiments — asymmetric outcomes, fast timelines, some things failing quickly and some scaling much faster than expected, intentionally or not. I’ve watched a project that started as a simple regex parser to pull terms out of contracts slowly expand, piece by piece, because once people saw what was possible, the scope just kept growing. That’s exactly why the ceiling and the checkpoint matter more here than they do for a normal software purchase — you’re not buying a fixed asset, you’re funding a bet, and bets need a defined size and a defined date to check the result.

Put the checkpoint on the calendar the moment the budget gets approved — not as a someday follow-up. I always tell people to use their calendar as a weapon: get that meeting scheduled weeks out, right when the green light happens, so nobody’s scrambling to find time later and the whole thing doesn’t quietly stall.

Key takeaways

• Every AI pilot pitch needs three things in writing: a named owner, a dollar ceiling, and a checkpoint date.
• “Let’s test it and see” isn’t an approvable request — it has no boundary and no accountability attached.
• AI spend behaves more like funding a fast experiment than buying a depreciating asset — budget and review it accordingly.
• Schedule the checkpoint the moment you approve the pilot, not as a someday follow-up — put it on the calendar immediately.
• A pilot that’s never actually reviewed against its original expectation has stopped being a pilot. It’s just become spend.

Want the CPE credit? Take the full lesson on EverydayCPE and earn 0.2 CPE credits: How to Pitch an AI Pilot and Actually Get the Budget