Estimated Read Time: 7 minutes
AdPipe is a creative analytics platform that uses computer vision to show why ads perform and helps teams scale what works. This article draws on AdPipe's Pulse Report 2026, a survey of marketing, creative, and agency leaders.
Pick one decision AI is supposed to improve, record a baseline for a metric you're already judged on, then compare the change against the full cost, including setup hours. In AdPipe's Pulse Report 2026, 78% of marketing teams couldn't point to clear, measurable results from AI, often because no baseline existed.
The basic formula is the same as any other investment:
AI ROI = (value of the improvement − total cost of AI) ÷ total cost of AI
The hard part isn't the math. It's agreeing on what "value" means before you start, and capturing a "before" number to compare against.
Most teams never set a baseline. In AdPipe's Pulse Report 2026, 64% of teams said AI delivers some value that's hard to measure, and another 7% weren't tracking it at all. The report's view is that "hard to measure" usually means "never baselined."
Feelings fill the gap where numbers should be. The Pulse Report found a 0.8-point gap between how teams feel about AI and what they say it has delivered. A wide gap like that means enthusiasm is doing the work that proof should be doing, which is fragile the moment a budget gets questioned.
For more on the proof gap, see Why Can't Marketers Prove AI Is Working?
Use a metric already on your scorecard. In AdPipe's Pulse Report 2026, 59% of marketing leaders named revenue or pipeline contribution as the metric that matters most this year, and teams that aimed AI at a metric they're measured on could prove results 40% of the time.
Respondents could pick more than one, so the column adds up past 100%. The report's advice is simple: don't invent an AI metric. Borrow the one you're already judged on, whether that's a rate, a cost, or a cycle time.
Not as the headline. In AdPipe's Pulse Report 2026, teams whose stated goal for AI was saving time proved results least often and ran fewer use cases. The report's reasoning: keep the hours as a bonus, but don't lead with them, because nobody can audit them.
Time saved only becomes ROI when it goes somewhere measurable. If AI frees up ten hours a week and those hours ship more tested creative, measure the creative. If the hours disappear into the week, there's nothing to count.
One respondent described what happens when time savings become the goal:
"When used correctly, it's an awesome tool. When used just to save time it creates slop and business none of us need." Creative and design manager, Pulse Report 2026
Want the full measurement playbook? Get the Pulse Report.
List the value of the change in your chosen metric, then list every cost, including setup and oversight time, and divide the net gain by the total cost. Only count the part of the change you can tie to AI, ideally by comparing against a baseline or a version that didn't use it.
Here's an illustration. Say a team uses AI to decide which creative to scale on $500,000 a month of paid social, and cost per result drops from $40 to $38.
These numbers are illustrative, not a benchmark. The point is the shape: a specific decision, a scorecard metric, a before-and-after comparison, and every cost on the table, including the hours.
Setup time belongs in the cost line on purpose. In the Pulse Report, lack of time or resources to implement was the single biggest blocker, named by 27% of leaders.
Monthly, even if the first numbers are rough. The Pulse Report's guidance is to report a rough number every month rather than wait for a perfect one, because precision improves as the habit sticks.
Start before anyone asks. The report recommends running the check while the program is still small enough to instrument, rather than waiting for a budget review to discover there's no answer.
Depth makes reporting easier over time. In the same survey, no team running only one or two AI use cases could prove results, while 43% of teams running seven or more could.
AdPipe is a creative analytics platform that uses computer vision to show why ads perform and helps teams scale what works. For creative, AdPipe ranks every creative choice by what it earned against your performance data, which gives teams a creative metric they can audit, like cost per result against the other ads in the same ad set.
That's one of the least-used ways to get value from AI. In AdPipe's Pulse Report 2026, only 12% of teams used AI to understand why their creative performs, and creative analysis and measurement took 5% of where next year's AI budget was headed.
See what your creative is returning. Get your free creative insights
The figures in this article come from AdPipe's Pulse Report 2026, a survey of marketing, creative, and agency leaders completed between July 7 and August 26, 2026. Results are unweighted. Respondents opted in to an AI survey, which likely over-represents engaged teams, and the report treats single-digit differences as noise.
The ROI example is a hypothetical illustration, not survey data. For the rest of the report's findings, see The AdPipe Pulse Report 2026: Six Findings on AI in Marketing.
How do you measure the ROI of AI in marketing?
Pick one decision AI is supposed to improve, record a baseline for a metric you're already judged on, then compare the change against the full cost of AI, including setup and oversight time. ROI equals the value of the improvement minus total cost, divided by total cost.
Why is the ROI of AI hard to measure?
Most teams never set a baseline. In AdPipe's Pulse Report 2026, 64% of teams said AI delivers value that's hard to measure, 7% weren't tracking it at all, and 78% couldn't point to clear, measurable results.
What metric should you use to measure AI ROI?
Use a metric already on your scorecard. In AdPipe's Pulse Report 2026, 59% of marketing leaders named revenue or pipeline contribution as the metric that matters most, and teams that aimed AI at a metric they're measured on proved results 40% of the time.
Should you measure AI ROI in time saved?
Not as the headline. Teams whose goal was saving time proved results least often in the Pulse Report, because hours are hard to audit. Time saved counts only when it goes into work you can measure.
What costs should be included in AI ROI?
Include the tool cost, the setup hours, and the ongoing oversight time. Setup time matters most: lack of time or resources to implement was the biggest blocker in the Pulse Report, named by 27% of leaders.
How often should you report AI ROI?
Monthly, even if the first numbers are rough. The Pulse Report recommends reporting a rough number every month rather than waiting for a perfect one.
What is AdPipe?
AdPipe is a creative analytics platform that uses computer vision to show why ads perform and helps teams scale what works. It ties each ad's creative choices to performance data, so teams can measure creative against a cost per result they can audit.