My board wants AI ROI in 90 days

Your board may have written "AI ROI," but what many boards are really buying in ninety days is a "we're in the race" story to tell—one of adult supervision: that management understands the opportunity, will not get embarrassed by a competitor story, and will not write a blank cheque. That is political and fundraising currency as much as it is financial math. Treat the quarter as a credibility window first. Sustainable returns sit behind a few clear catch phrases and some transparent learning that shows you are moving in the right direction.

Once "AI progress" is on the scorecard, every department will discover a favourite idea. Sales wants better proposals. Marketing wants more content. Ops wants automation. IT wants a platform. Legal wants to slow everyone down and put disclaimers everywhere. The leaders who proposed the AI investment may need a visible win. You may be the one expected to validate savings that nobody is willing to remove from a budget. If you do not put a decision process around this early, you will fund several experiments and still have to explain why nothing meaningful shipped.

Start with one expensive, repeatable problem that has a low cost of failure. Work that happens often, already has an owner, and already burns time or money you can measure. Write the baseline before anyone builds: hours, loaded cost, error rate, cycle time, cost of delay. If you cannot describe today's cost, you cannot defend tomorrow's ROI without inventing it.

Then pick one workflow that can survive IT, legal, and the people who do the work. Give it a sponsor, an operational owner, and someone who owns the number. Keep it narrow enough to learn from, real enough that someone notices. A clever demo is not a changed workflow. By day ninety, report three things honestly: what changed, what you learned, and what you recommend next—including what you will stop funding. A well-governed "no" can be more valuable than a pilot kept alive to protect someone's reputation.

Watch the human dynamics: the team executing may quietly fear that efficiency means fewer roles. Technology may inherit maintenance without capacity. The executive who wanted AI on the deck may care more about the story than the number. Bring those incentives into the room when you choose the first use case, or everyone will celebrate the pilot and quietly reject the change required to realize it.

Your board update should not be "we explored AI." It should be: "We selected one costly workflow, established a baseline, made a controlled change, and now have evidence to stop, improve, or scale." That is credible progress. The goal is to show the organization can make disciplined decisions about AI and that you will not sign for theater. Say it from the outset of the engagement and make sure a KPI is to document the learning journey along the way. Done properly, the company has the "win" even if the project is not a success—because everyone was able to learn more along the way.