That project got approved anyway, paid for itself within a couple of months, and turned the skeptics into people asking what to automate next. The approval came down to framing: three specific moves worth copying.
How do you frame AI spending for a skeptical board?
As revenue with a monthly off-switch, not as a capital project. The CTO above didn't pitch "AI transformation." He pitched a specific automation of a specific high-margin revenue line: spare-parts quoting that was bottlenecked by manual work, with a measurable backlog and an obvious payoff.
Then he framed the commitment: month-to-month, continuing only while it delivers value. That converts the board's decision from "do we believe in AI," a question a skeptical board answers no, into "is this month's value worth this month's cost," a question the results answer for them. It also keeps the vendor honest: the engagement only keeps going if value keeps happening.
What's the alternative cost the board should compare against?
A full-time AI hire, and that comparison is what closes the argument. A dedicated AI specialist runs well north of $200,000 a year fully loaded, takes months to find and ramp, is genuinely hard for a manufacturer to evaluate, and might not work out. For a company watching profitability, that's a heavy, slow, illiquid bet on an unproven role.
Set against that, a month-to-month engagement that's already producing isn't the risky option. It's the conservative one. That reframing, not cost versus zero but cost versus the realistic alternative, is what moved a board that was openly skeptical of AI.
What if IT constraints are out of your control?
Work what's unblocked and push weekly on what isn't. The CTO above didn't control internal IT; the parent company did. Some permissions came through, others sat for months: granted, then ungranted, then granted again. The project moved anyway, because the team kept progressing on available infrastructure in parallel while maintaining a steady weekly cadence on outstanding access.
The architecture helped. A system that deploys in your environment fits itself to the infrastructure you can actually get, rather than requiring the perfect setup before anything starts. Waiting for ideal conditions is how these projects die in committee; the working posture is that there's never a reason to fully stop.
How do skeptics actually come around?
They see a result, not a better pitch. No deck moved this board. What moved them was the first project landing: the bottlenecked revenue line unblocked, the payback visible inside a quarter. After that, the conversation flipped from whether to do AI to what to do next with it.
That's the general lesson for anyone navigating a skeptical leadership team. Don't argue the category: manufacture one undeniable, measured win and let it argue for you. Pick the first project for provability: clear before-state, countable backlog, revenue or hours you can measure on the other side.
What does a board-ready first proposal look like?
One page, five elements. The specific process being automated and its current cost in hours or backlog. The revenue or savings it gates. The commitment terms: month-to-month, value-contingent. The comparison cost of the realistic alternative (a specialist hire, a bigger software project). Where it runs: in your environment, with your data staying in your building, which preempts the security objection before it's raised.
A board can say yes to that without saying yes to "AI." Which is exactly the point.