AI in marketing: the board wants ROAS, not hours saved

You have seen the slide. "AI saves us 12 hours a week." Everyone nods. Then the CFO asks what those 12 hours earned. Silence. A survey from US software company Validity, published November 11, explains why that silence is about to get expensive.

What the survey found

• Pressure from the top: 61 percent of marketers surveyed feel pressure from the C-suite to show measurable AI gains.
• What executives want: as leadership's top metric, 25 percent name marketing ROI or ROAS, 19 percent revenue and pipeline contribution, 16 percent customer acquisition cost. Only 6 percent name efficiency.
• What teams measure: those reporting a measurable AI return point mostly to cost savings (41 percent), engagement (39), time saved (35) and conversions (35). Only 26 percent measure revenue against AI cost.
• The 2026 scorecard: half say their AI investments met or beat ROI expectations. 26 percent fell short or saw no measurable return, 24 percent cannot tell yet.
• 2027: 53 percent plan to put at least 10 percent of their budget into AI. Only 7 percent let agents run workflows with minimal human oversight.

Time is not a currency in the boardroom

To me, that is the real finding: teams measure what is easy to measure, executives buy what shows up in the P&L. An hour saved only becomes money when it generates revenue somewhere else or replaces a hire, an agency invoice or a tool. Until the report says that, efficiency is a claim.

My rule of thumb: AI is the ticket in, not the job. No prompt decides where the budget has the most leverage. That decision is exactly what the board wants to see. Showing output volume, more posts, more variants, faster briefs, answers a question nobody asked.

The catch: cuts follow the math

One detail should make every team nervous. Asked what they would protect first in a mid-year cut, 20 percent pick paid media and only 16 percent AI tools and agents. If you cannot translate your AI stack into revenue, it is on the cut list. And AI does not fix broken ways of working. If tracking and attribution were shaky before, the AI report does not get cleaner, just longer.

For context: Validity is a US software vendor, the sample is just over 250 people, and the release names neither countries nor fieldwork dates. Treat it as a signal, not proof. But anyone who has defended a budget knows the logic.

What I would do now

• → Tie every AI use case to exactly one business number: ROAS, pipeline, acquisition cost or external spend saved. No use case without a number.
• → Convert saved hours: where did the time go and what did it produce there? If the answer is "more meetings," that tells you something too.
• → Keep final approval and budget allocation in-house. That is the part AI does not take over, and the part the board asks about.

Bottom line: the board does not pay for hours saved, it pays for what you earned with them.