Almost every marketing team is using AI now. New research from Alexander Group found adoption has reached 96%, but only 41% of marketers can prove that AI is producing a return.
The ROI gap is only part of the story. While 56% of companies use advanced attribution models, only 30% trust the accuracy of those results across Marketing and Sales.
Proving AI ROI becomes much harder when the underlying measurement is already in question. If teams already disagree about which campaigns, channels, and interactions deserve credit for revenue, adding AI to the process gives them another investment to measure with numbers they weren’t completely confident in to begin with.
Why AI ROI Is Hard to Measure
A new ad platform at least gives marketers a defined investment to evaluate. They can track spending, campaign performance, and the opportunities associated with that channel, even if determining its exact contribution to revenue remains difficult.
AI rarely stays inside one clean box. It can help research an audience, write a campaign brief, create content, prioritize accounts, summarize sales conversations, personalize outreach, and forecast pipeline. One AI tool can influence several parts of the buyer journey without owning any one of them.
It can be hard to tell exactly how much AI contributed. A campaign might perform better after a team introduces AI. The improvement could come from better targeting, stronger creative, or something else that changed along the way. It's not easy to know how much credit AI deserves.
Alexander Group’s research shows this against an already complicated buyer journey. Buyers now complete about 70% of their evaluations across 10 channels before engaging directly with a company, while customer acquisition costs have increased at least 60% since 2022.
Why Marketing Attribution Still Has a Trust Problem
Marketing has spent years adding more measurement tools in an attempt to explain where revenue comes from. Multi-touch attribution, intent data, CRM reporting, and platform analytics can all provide useful information, but more models don’t automatically create agreement about what happened.
The Alexander Group finding makes that clear. More than half of the companies surveyed already use advanced attribution, while fewer than a third trust the results across Marketing and Sales.
Of course, disagreement doesn't always mean the data is wrong. Marketing and Sales may be measuring different parts of the same journey.
But when teams can't agree on what the numbers mean, those numbers become a shaky foundation for proving another investment's value.
The lack of trust becomes a problem when teams start making AI ROI claims. If Marketing credits one set of activities while Sales believes something else actually moved the deal, adding an AI-assisted touchpoint to the attribution model can make the argument more complicated rather than resolving it. Before looking for a better AI dashboard, teams need to agree on what counts as success.
Agree on the Evidence First
AI becomes easier to measure when Marketing and Sales agree on the job before the work starts.
Say a demand generation team introduces AI to prioritize accounts. Six months later, its pipeline has grown 20%. That's encouraging, but it doesn't tell the team whether AI caused the improvement. The team would need to look at whether those accounts converted at a higher rate, moved through the pipeline faster, or simply received more outreach. Those comparisons would help show how much of the improvement actually came from AI.
It works the same way for other uses of AI. For campaign production, faster drafts don't mean much if campaigns still take just as long to launch. Media efficiency works the same way, but the bar is higher. There, teams need to look at costs, qualified demand, and revenue to see whether AI actually improved performance.
That agreement has to happen before the test begins, and it has to include the people who will be asked to believe the result. Otherwise, teams can end up searching through dozens of metrics afterward for one that makes the investment look successful.
Alexander Group found that only 31% of marketers have established managed AI governance. Its research also found that top performers are applying AI across more use cases while proving contribution through productivity gains.
It comes down to accountability. Marketing and Sales need to agree on what AI is supposed to improve and how they'll measure whether it worked.
Better Decisions Need Proof
At RAD Intel, we think the evidence behind a marketing decision matters as much as the recommendation itself. Audience intelligence can help marketers understand why a particular audience, creator, or strategy makes sense before committing resources. But the process can't end there. Teams also need to compare what actually happened against what they expected, so they can determine whether the decision delivered value.
Measurement needs to be part of the decision from the start. A team should know what it expects to learn, which business result matters, and how it will compare the outcome before money is committed.
Most marketing teams are already using AI. Now they need to prove it's making a difference, using numbers Marketing and Sales can both trust. Until that happens, the AI ROI problem will remain tied to a much older marketing problem. Everyone has a number, but not everyone believes the same one.
