Proving the return on AI: measure the approved work, not the draft
Marketers have always had to defend their budgets in the language of finance. AI has made that harder, not easier.
The spending is no longer small. Gartner's 2026 CMO Spend Survey of 401 marketing leaders in North America, the UK and Europe found CMOs allocate an average of 15.3% of marketing budgets to AI initiatives. Only 30% report mature or fully developed AI readiness. Meanwhile, The Martech Weekly's 2026 outlook, reported by MarTech in October 2026, found 40.2% of martech leaders still can't show a clear financial contribution from their technology investments.
That is a professional problem as much as a technical one. If we cannot show what AI returns, the finance director will draw their own conclusions.
Why the usual numbers mislead
Most AI reporting counts activity: drafts produced, prompts run, seats bought. None of it tells a CFO whether the business is better off. Worse, it measures AI at the point where it looks best.
Optimizely's June 2026 study of 2,003 marketing leaders across seven markets, including the UK, found only 4% say AI saves them time at every stage of the process. The Content Marketing Institute found the same gap from another angle: 87% of B2B marketers using AI for content said productivity had improved, but only 39% said content performance had. The draft got faster; the results often did not.
Measure the whole cycle
The answer is to measure from brief to approved asset, before and after AI. Six measures cover it:
- Cycle time, from an agreed brief to approval.
- Touch time, the total time people spend on the asset.
- Checking time, the share spent reviewing and correcting AI output.
- Rework rate, split into context fixes and polish fixes.
- Approval wait, the time an asset sits in a queue.
- Cost per approved asset.
The split in point four is the one to watch. Polish fixes are ordinary editing. Context fixes mean the AI lacked something your team knows: the audience, the claim you can stand behind, the tone. If context fixes dominate, adding more AI will lengthen the queue.
Take a baseline first
Log one normal working week on five recurring assets before AI changes the workflow. Record who touched each one, for how long and how long it waited. Then repeat the log on the same assets once AI is in place. Without a baseline, every improvement gets credited to AI, including the ones that came from a new process.
Speak the CFO's language
Finance will accept measures that trace to money. Those are a lower cost per approved asset, supplier spend avoided, time released and redeployed to named work, and incremental revenue proven against a control group. Time saved counts only when it is spent on something else of value.
Set the stopping rule before you start, too. Gartner's survey of 2,002 CIOs and technology executives, published in October 2026, found only 31% of CIOs had guidelines for stopping an AI pilot before production.
Why this matters for the profession
Proof changes outcomes. In The Martech Weekly research, 56% of teams that could prove value received budget increases, against 37.5% of those relying on what the report calls "faith-based" value demonstration.
For CIM members, this is familiar ground. We already know how to set objectives, take a baseline and evaluate the result. AI does not change that discipline. It makes it urgent, because the cost of checking AI output is real, and it belongs in the business case beside the licence.
The full guide, with a CFO scorecard and what an AI business case for 2027 should contain, is on the Jam 7 blog: AI marketing ROI: how to prove AI saves time and money.
About the Author
Jack Hardy is the Chief Marketing Officer at Jam 7, an award-winning Chartered Marketer, and a CIM board member with over a decade of experience in crafting B2B growth strategies across various technology platforms and SaaS providers.