This article is based on a Q&A with BrainTrust Partners CEO Tim Hayden and Greg Kihlström for The Agile Brand Guide’s Expert Mode series.
Most of us have settled on a story about AI spending by now. Budgets went up, pilots multiplied, and the returns never quite arrived. The numbers seem to agree. KPMG’s Q1 2026 Global AI Pulse, which surveyed 2,110 senior leaders, found that 95% of organizations have an AI strategy and only 8% report established ROI.
But the same KPMG survey found 64% reporting meaningful business value. And Dun & Bradstreet’s AI Momentum survey of 10,000 businesses (research from a company that sells business data, which is worth keeping in mind) found more than three-quarters of enterprises seeing some measurable return, while just 6% called their data fully ready to support AI at scale. So value is showing up. What isn’t showing up is proof a CFO will sign off on. Tim Hayden, CEO of BrainTrust Partners, an Austin consultancy that builds AI roadmaps and customer data systems (no relation to the AI observability company Braintrust), argues that marketing keeps measuring the wrong thing. He has a stake here. BrainTrust is launching an AI assessment and a data platform of its own, so we asked him to argue against both.
“Established ROI” Is a Retention Number
Asked which of the two survey figures matches what he actually sees inside marketing organizations, Hayden didn’t pick one. He moved the goalposts instead, and arguably to a better spot.
The reality is that AI is present and producing measurable changes in workflows, yet no one is sharing through earnings reports, layoffs, or store closures how AI is truly impacting their bottom line KPIs. “Established ROI” is left to those companies who are lowering their COGS/opex while growing revenue with minimal churn in employees or repeat customers. CMOs need to show they are growing Customer Lifetime Value (CLV), not simply filling the top of the funnel, while lowering the operating costs and reducing their media dependencies.
That’s a strict definition, and most marketing AI business cases fail it. They’re built on activity: hours saved, assets produced. Those gains are real. They’re also inputs, and finance knows it.
For a CMO, the practical translation is to tie every AI line item to two things before it gets approved. The first is a customer lifetime value cohort it’s supposed to move. The second is a media line it’s supposed to shrink. If an AI-assisted retention program lets you pull back paid acquisition and hold new-customer volume steady, that’s a number the CFO recognizes. If the case rests on content velocity, expect it to land with the 64%. Valuable, sure. Not established.
A Diagnostic Is Only as Good as the Room
BrainTrust’s AI BluePrint, one of the two products launching September 22, is a guided consultation the company says takes about an hour. Hayden describes it starting with company size, team titles and responsibilities, and existing technology, then moving into workflows, how long tasks take, and how reports get made and decisions get made. It ends with prioritized use cases. We asked where a marketing leader would get a confidently wrong answer out of it.
A marketing leader would get a confidently wrong answer out of it if participating in the consultation alone, or if sharing unqualified information during the consultation. The BluePrint will be defensible only if the consultation is informed and thoughtful.
That warning applies to every AI readiness assessment, including the six-week versions from large consultancies. The failure isn’t the tool. It’s a marketing leader describing workflows the way the org chart says they run rather than the way they actually run. The monthly performance report that “takes a day” usually takes three, spread across two analysts and a spreadsheet nobody else understands. An assessment built on the org-chart version will prioritize the wrong use cases with total confidence.
So staff the session like it matters. Bring the marketing ops lead who knows how long things really take, someone from IT who knows where the customer data lives, and ideally someone from finance, who’ll have to believe the ROI model later anyway.
One more thing. We also asked Hayden to make the case for the CMO who should skip the assessment entirely and just run a few pilots to learn from what breaks. He declined. “We no longer encourage experimentation without diligence or an assessment, as AI risks are largely untaught and not respected,” he said. It’s a defensible position, and it’s also the position of someone who sells assessments. The honest middle ground is that some diligence should come before pilots. How much is the real question, and your answer and a vendor’s may reasonably differ.
Where Agents Actually Break
Data readiness tends to stay abstract until someone names a failure. Hayden named a few, starting with data that’s unlabeled and hard to audit. He also argued that most businesses can’t honor a customer’s request to correct their data, pause its processing, or delete it outright, as state privacy and AI governance laws increasingly require. That’s his assessment rather than a measured figure, but it’s one any CMO can check. Then he offered the example that actually lands.
Every day, companies are redundantly communicating or operating the same tasks multiple times without connected data orchestration. Try tweeting a complaint to American Airlines two minutes before your turn at the counter, while having your wife change the flight on the AA mobile app.
Notice what that example describes. It isn’t an architecture diagram. It’s one customer showing up in three channels at once, and none of those channels knowing about the others. Put AI agents on top of that setup and they won’t fix the confusion. They’ll just move through it faster.
Hayden’s point about privacy requests hands CMOs a test they can run in a week rather than a claim they have to take on faith. Submit a deletion request as a test customer, then trace how many systems it actually reaches and how long that takes. Separately, pick a customer who complained on social media last month and check whether your CRM and email platform knew about it. Both exercises are really tests of identity resolution. If either one takes more than a few days to answer, your data isn’t ready for an agent, whatever the vendor demo suggested.
The Unglamorous Fix
Asked for the single change a CMO should make with one quarter left and no new headcount, Hayden’s answer doesn’t require his company’s product. That’s worth crediting.
The CMO invests in data unification and resolution with a CDW managed with Snowflake, and licenses a marketing automation system to better orchestrate email and SMS/RCS communications with current/known customers. Better audience insights, agentic reporting and alerts on outreach or content opportunities, and direct 1:1 communications with a market that already shows trust in your brand.
It’s the same argument he made about ROI, expressed as a spending decision: shift from acquisition toward customers you already have. Hayden projects that return and repeat engagement can double within a quarter or two, with referrals from known customers adding what he called “gravy” to the funnel. Treat that as a hypothesis to test, not a benchmark. Still, the direction is sound.
For most enterprise marketing teams, the licensing part is probably already done. You likely own a data warehouse and some form of marketing automation. The work is connecting them so the warehouse’s view of a customer actually reaches the email and SMS programs in something close to real time.
Hayden didn’t say what to stop funding to pay for it, so here’s a starting point. Look at the share of paid media that’s retargeting people who are already in your CRM. You’re paying a platform to reach a customer you could message directly. That’s the media dependency from his first answer, and it’s usually the easiest money to move inside the same budget cycle.
The 8% figure isn’t really evidence that AI doesn’t work. It’s evidence that most organizations haven’t decided what would count as working. Hayden’s definition (growing lifetime value, lower operating cost, less reliance on paid media) is one a marketing leader can write down before the next budget meeting. So write it down. Before you approve the next AI line item, name the customer cohort it should move and the media line it should shrink. If nobody in the room can name them, that’s your assessment.