A survey of marketing chiefs around the world just put hard numbers on something you've probably felt in your own business: everyone is spending more on AI, and almost no one can tell you what it's earning.
The report — Comviva's Global CMO Survey, fittingly titled "The AI Efficiency Divide" — found that 90% of organizations increased their AI marketing investment over the past two years. Near-universal adoption, across every region and sector. And then the other number: only 12% can rigorously isolate what that AI actually added in revenue. The rest are working off vibes, proxies, or nothing at all.
It gets more specific. 67% of these companies can't even total what they spend on AI, because the costs are scattered across cloud bills, tools, vendors, and people. 79% rely on estimates rather than real measurement. And 86% of leadership teams are now demanding proof — which means a lot of marketers are about to be asked a question they can't answer.
Why this matters if you're not a CMO
These are big-company numbers, but the pattern shows up at every size. If you run a small business or a small marketing team, the miniature version looks like this: a ChatGPT subscription, a Canva upgrade, an AI feature inside your email tool, maybe an AI scheduler — each one reasonable on its own, none of them tied to a number. Ask yourself the leadership question: what did AI earn you last month? If the honest answer is "it saves me time, I think," you're in the 88%.
Here's the part worth sitting with: the companies in the 12% aren't there because they bought better AI. They're there because they measured before they scaled. That's not a technology advantage. It's a habit — and habits are free.
The 12% didn't buy better AI. They picked one number per workflow and watched it. That's the whole trick.
How to be in the 12%, at your scale
You don't need attribution software or a data team. You need one number per AI workflow, written down before you automate anything.
Pick the workflow first, the number second. Say it's follow-up emails to leads. The number is reply rate, or booked calls. Not "engagement." A number you'd bet on.
Get a two-week baseline. Before the AI touches it, count what the workflow produces now. Two weeks of honest counting beats a year of guessing. Most businesses skip this step, which is exactly why they can never answer the ROI question later — there's nothing to compare against.
Then automate, and compare. Same number, same two weeks, AI in the loop. If replies went from 9 to 14, you have proof. If nothing moved, you just saved yourself a subscription — that's a win too. Kill it and try the next workflow.
Total your AI costs once a month. The 67% who can't sum their spend aren't careless; the costs just hide in different line items. Yours fit on a sticky note: every tool, every AI add-on, one monthly total. Now the two numbers — what it costs, what it moved — sit next to each other, and the ROI question answers itself.
The uncomfortable takeaway
The report's real message isn't that AI in marketing doesn't work. It's that spending has been running about two years ahead of measuring, and the bill for that gap is coming due — 86% of leadership teams are already asking. The businesses that can answer will keep their budgets and double down on what's working. The ones that can't will cut tools at random, including the ones quietly earning their keep.
Being small is an advantage here. A CMO has to untangle AI spend across forty tools and six departments. You have four subscriptions and one workflow that matters most. You can be in the 12% by Friday.