Marketing Poured 15.3% of Its Budget Into AI. Most Companies Still Can't Find the Return.

On September 30, 2025, Boston Consulting Group published a survey of 1,250 senior executives across more than 25 sectors. The headline number was small. Only 5% of companies were pulling substantial value out of their AI investments. About 60% reported minimal revenue or cost gains, despite spending real money to get there.

Marketing is spending that money faster than almost any other function.

Gartner’s 2026 CMO Spend Survey, fielded January through March, found CMOs now put an average of 15.3% of their marketing budgets into AI, while only 30% of marketing organizations say they are ready to scale it. One in every six or seven marketing dollars is going toward a capability most teams admit they cannot yet run. The invoices are current. The payback is not.

That gap is now the most consistent finding in the 2025-2026 data. Different research houses, different samples, same shape: money in is up and to the right, and provable return is a flat line a few quarters behind it.

Nobody in this data is failing to spend. They are failing to prove the spend worked.

Start with the widest read. McKinsey’s 2025 State of AI survey found 88% of companies now use AI regularly. Only 39% report a positive impact on earnings from it, a figure the Content Marketing Institute put in front of marketers directly this year. Adoption is near universal. Bottom-line impact is a minority event.

Marketing’s own numbers say the same thing. In Jasper’s State of AI in Marketing 2025, only 49% of marketers said they measure the ROI of their AI investments at all. You cannot report a return you never instrumented. And when BCG asked CMOs how they were prioritizing spend on the customer experience and the martech stack, measurement ranked last on the list.

So the picture composites cleanly. Marketing is among the fastest adopters and the fastest spenders. It is also among the least equipped to prove what the spend bought.

The mechanism is not mysterious. A tool subscription clears procurement in a week. The data plumbing, attribution, and process redesign that turn a subscription into measurable pipeline take a year, and most teams bought the first without funding the second. BCG’s own read is that the 5% who capture value are the ones who reshaped how the business works, not the ones who bolted AI onto the org chart they already had. The 60% getting almost nothing back are running new tools inside the old operating model. That is exactly where marketing sits when it hands a content team a chatbot and calls it a strategy.

There is a second reason the return lags. Marketing budgets did not grow to make room for AI, so the AI line is a reallocation. Something got cut to pay for it. Often it was people: 47% of B2B SaaS companies cut or stopped backfilling marketing roles because of AI in the past year, per Wynter’s survey. When headcount funds the tool, the return has to clear a higher bar. It is not enough for AI to help. It has to help more than the person it replaced.

This is fine while AI is a story. It stops being fine the moment it becomes a line item with a history. Finance treats any spend that has been around long enough to have a track record the same way it treats every other one: prove it or lose it. A budget category worth one-sixth of marketing spend, attached to a capability two-thirds of teams say they cannot scale, is the first thing a CFO circles when the quarter tightens. The CMO who cannot connect that spend to pipeline is the one holding it when the question comes.

The uncomfortable part is that the leaders are pulling away while the laggards keep spending. BCG found its top performers get roughly five times the revenue lift and three times the cost reduction from AI that everyone else gets. The distance between spending on AI and earning from it is not closing on its own. It is widening, and the companies on the wrong side of it are still writing the same checks.

Three moves separate the 5% from the 60%. Instrument first: no AI tool renews without a measurement plan tied to pipeline, because the half of marketers who do not measure are budgeting on faith. Fund the substrate, not just the subscription, because the data and attribution work is where the return actually lives, and it is the line CMOs keep cutting to afford more tools. And give the whole AI allocation a single owner with a number attached, before the CFO assigns one for them. The spend is already committed. The only open question is who gets asked to defend it, and when.