Forrester just flagged it: enterprises are expected to delay 25% of planned AI spend in 2026 as ROI stays elusive. That means Q2 and Q3 board conversations are about to get uncomfortable for a lot of technology leaders.
Here’s the thing most of those leaders don’t realize: the companies that survive budget scrutiny aren’t the ones with the best results. They’re the ones who built measurement into their projects from day one.
Everyone else is walking into that meeting with vibes.
The Measurement Gap
The reason AI ROI conversations go badly isn’t that AI doesn’t work. It’s that most implementations were never instrumented to prove that they work.
The pattern: a team gets excited about a use case, builds something, ships it, and then six months later someone asks how it’s performing. The honest answer is usually “we think it’s helping, but we don’t have numbers.” That’s not an answer a board accepts when they’re looking for reasons to cut spend.
The projects that survive scrutiny defined their success metrics before they built anything. Not after, when you’re trying to retrofit measurement onto a system that wasn’t designed to be measured.
What to Bring to the Meeting
If you’re preparing for a board conversation about AI ROI, here’s the actual checklist:
A baseline. What was the process before AI? How long did it take? What did it cost? What was the error rate? If you don’t have this, your ROI calculation is a guess. The board will know it’s a guess.
A specific metric that changed. Not “we improved efficiency” — a number. Processing time went from 4 hours to 40 minutes. Support ticket volume dropped 23%. Document review cost per unit fell by 35%. One specific metric with a before and after.
The measurement methodology. How did you measure it? Was it a controlled comparison? A time-series before and after deployment? An A/B test? Boards that are doing their job will ask. “We just looked at the numbers” is not a methodology.
The cost of the AI layer. API costs, infrastructure, engineering time, ongoing maintenance. ROI is a ratio. You need both sides of it. A lot of AI projects report the top-line benefit without the full cost — that’s the math that falls apart under questioning.
What you’re not measuring yet. This one is counterintuitive, but being honest about what you don’t know builds more credibility than overstating what you do. If you’re tracking cost reduction but not quality impact, say so. If you haven’t fully attributed the time savings, say so. Boards trust leaders who know the limits of their data.
The Real Problem Isn’t the AI
The companies facing budget scrutiny right now usually don’t have an AI problem. They have a measurement problem that predates AI.
They ran projects the way they always ran projects: build something, ship it, see if it seems to help. That approach works fine when you’re not being asked to justify the spend to a board. It doesn’t work when every line item is under pressure.
The fix isn’t complicated. Before you start the next AI project, write down what success looks like in measurable terms. Set a baseline. Instrument the output. Plan to run a comparison. None of that is hard — it just has to happen before you build, not after.
The board conversation is the forcing function. The real work is the instrumentation you put in place twelve months earlier.
VitaLink Software helps enterprises build AI systems with measurement built in from the start — not bolted on when the board starts asking questions. If you’re heading into a budget conversation you’re not prepared for, that’s where we start.