How to build an AI business case that survives a steering committee

AI optimism is not a business case. The language around AI initiatives keeps climbing: transformation, automation, productivity, intelligence, scale, competitive advantage. None of those words survive contact with a steering committee that has funded three AI pilots this year and cannot point to a return on any of them.

An AI initiative still succeeds or fails on project fundamentals, the same ones that have always decided whether a business case gets funded. The language around it has just gotten better at hiding that fact.

If the case in front of you only says "we need AI," it is not ready for investment, no matter how compelling the demo was. Before you take it to the room, be able to answer seven questions in plain business language, not in model terminology.

What problem actually changes if this works. Not what capability gets added, what problem stops being a problem.

What work changes, specifically, for specific people, not "productivity improves."

What data is required, and do you actually have it, cleanly, with the rights to use it.

Who trusts the output, and what happens the first time the output is wrong.

What risk is introduced that was not there before: model dependency, data exposure, vendor lock-in, regulatory exposure.

What value proves success, defined before the pilot starts, not interpreted afterward to fit whatever happened.

Who remains accountable when the system makes a bad call, because "the AI decided" is not an accountable answer in any steering committee that has been burned before.

PMs should not resist AI enthusiasm in the room. It is genuine energy, and killing it is a poor use of political capital. The job is to direct it, not dampen it. Every AI pitch that walks in as raw enthusiasm has to leave the room as an investable ask, meaning a defined problem, a measured pilot, a named accountability owner, and a pre-agreed definition of what counts as evidence that it worked.

Adoption data backs the urgency here. Reported AI use on projects rose from 36 percent in 2023 to 70 percent in 2025, and the tooling is genuinely outpacing the governance sitting underneath it in most organisations. That gap is exactly where a weak business case gets funded on enthusiasm and quietly fails eight months later with nobody able to explain why.

The business cases that survive scrutiny are rarely the ones with the biggest scale ambition in the first slide. They are the ones that can show a small, honest pilot result: this specific workflow, this specific team, this measured outcome, these specific limitations already found and already being addressed before asking for more budget.

A steering committee that has seen AI pitches before is not looking for confidence. It is looking for evidence that someone has already found where this breaks, on a small scale, before asking to fund the version where it breaks at enterprise scale.

Work through this before the case goes to committee.

Problem statement. Is the problem defined in business terms a non-technical sponsor could repeat back, not in model or feature terms?

Data readiness. Do you already have the data this needs, with the rights and quality to use it, or is that itself still an open risk?

Evidence base. What pilot result, however small, backs the ask, and what did that pilot already teach you about where this breaks?

Success definition. Is the metric that proves success written down and agreed before the pilot starts, not chosen afterward to fit the result?

Accountability. Who is named as accountable for the outcome, including the outcome where the system gets it wrong?

Risk register. Have you named the new risks this specifically introduces, not just the ones every project already carries?

Take the AI initiative currently sitting in your pipeline and write one paragraph answering each of the seven questions above, in language a sponsor with no technical background could repeat back. If you cannot answer two or more of them yet, that is not a reason to shelve the idea. It is the actual pre-work the business case needs before it goes anywhere near a steering committee. PM Strategy Advisor is built for exactly this moment, turning AI ambition into the governed, defensible ask you can walk into the room with.