Enterprise Marketing AI Is Stuck. Challenger Brands Built the Workaround.

Mike Shields cornered two enterprise CMOs at a recent industry dinner and asked them the question every vendor deck dodges. What is your real return on the marketing AI you bought? The answer, from both: basically nothing.

Same week, 55 Orangetheory Fitness studios across 10 states were running Passionfruit AI in production. Hourly lead tracking. Real-time media mix optimization. The analyst team that used to do this work is gone. Not augmented. Replaced.

Two data points. Same quarter. Opposite outcomes. If you read this as a tooling difference, you will misallocate the next budget cycle. The challenger brands are not winning because their AI is better. They are winning because they have less organization standing between the model and the decision.

The “basically nothing” admission

Shields, who covered ad tech at the Wall Street Journal and Business Insider before launching Next in Media, is not a vendor critic by trade. When two enterprise marketing leaders tell him separately that their AI returns are negligible, the signal is not a one-off complaint. It is what enterprise marketing AI looks like when you strip out the case study selection bias.

Read the failure mode carefully. These CMOs did not say the models do not work. They did not say the vendors lied. They said the ROI is basically nothing. That is the language of capability without realization. The technology is doing something. The organization is absorbing the value before it reaches the P&L.

This is the realization problem dressed in a marketing suit. The model produces a recommendation. The recommendation has to clear brand. Then legal. Then a regional approval. Then a global media review. By the time the recommendation acts on a campaign, the moment has moved, the budget cycle has closed, and the decision has been diluted into a committee compromise.

The vendor sold capability. The org chart consumed it.

The Orangetheory contrast

Alan Magee, CMO of Empire Portfolio Group, runs marketing for 55 Orangetheory studios. That is not a small business. It is a multi-state operation with real budget, real complexity, and real customer data. He gave Passionfruit AI a live job: aggregate the lead data, optimize the media mix, run it hourly.

Before Passionfruit, this work required a dedicated analyst team. After Passionfruit, the analyst team is not part of the workflow. The AI ingests the lead data, attributes spend, and surfaces the optimization. The CMO looks at the output and adjusts.

Raffi Salama, CEO of Passionfruit, framed it for Shields: “It’s the smaller brands that will compete with titans in ways they never could before.”

Salama is right about the direction and wrong about the cause. The smaller brands are not competing because the tool is small-brand-shaped. They are competing because the decision path between AI output and budget action is short enough that the AI can actually change the spend before the spend has already happened.

A 55-studio chain has one CMO, one budget owner, and one approval. The enterprise equivalent has eight CMO-equivalents, four matrixed budget owners, and a brand committee that meets every other Tuesday. Same AI. Same data. Different organization. Different outcome.

Where governance becomes the brake

The standard reading of enterprise marketing AI failure blames the platforms. Meta’s AI connector ships with no granular permission control. Performance Max is a black box. The vendor stack is fragmented. All of that is true. None of it is the binding constraint.

The binding constraint is who has to sign before the AI’s output becomes an action.

In the Orangetheory case, the answer is one person. In the enterprise case, the answer is a workflow. The workflow exists because the enterprise has more brand surface, more regulatory exposure, more historical accidents that produced new approval gates. Each gate was rational when it was installed. Every gate together produces an organization that cannot operate the technology it bought.

This is not a tooling debate. It is a governance design choice that no one made deliberately. The approval chain grew by accretion. The AI walked into the chain expecting to be a participant and discovered it is a recommender to a recommender to a recommender. The capability never reaches the spend decision in time to change it.

The diagnostic the CMOs are not running

If your AI ROI is basically nothing and your vendor’s case studies show 10x results at smaller companies, the honest question is not “which AI should we buy next?” It is “what would have to be true about our organization for this AI to produce value?”

Three tests, in order. Each one is answerable in a week.

First, time from AI recommendation to budget action. Pick one campaign. Trace the path. How many people touched it? How many days elapsed? If the answer is more than two people and more than 48 hours, the AI’s optimization signal is stale before it lands. The model’s edge is in cadence. You bought a system that runs hourly and you are deploying it into a workflow that runs quarterly.

Second, where the approval gate adds value the AI did not already address. Most enterprise marketing approval chains were designed before the AI could explain its own recommendation. The brand reviewer was checking what the agency produced. If the AI now produces the recommendation with the brand constraints encoded, the gate is reviewing a problem the system already solved. Document the value each gate adds. If a gate cannot point to a decision it has changed in the last quarter, that gate is org chart, not governance.

Third, who owns the loss when the AI does nothing. This is the question that surfaces the real problem. Enterprise marketing teams have spent two years buying AI and reporting investment. No one is on the line for the realization. The CMOs in Shields’ dinner did not say the basically-nothing ROI is showing up in their performance review. The investment is reported. The realization is invisible. The org chart has no row for “AI value not captured.”

What the challenger brands actually have

Orangetheory has 55 studios. They do not have an AI strategy. They have an AI in production. The difference is not branding. It is operational: the path from “model says spend more on Meta in Tampa this week” to “Meta budget shifted in Tampa this week” is short enough that the model’s recommendation is still relevant when the action happens.

Enterprise marketing teams will not get this by buying a better AI. They will get it by deleting approval steps that no longer add value. That is not a vendor decision. It is a leadership decision. The CIO did not install those gates. The CMO inherited them.

The honest version of the Salama quote is this: smaller brands compete with titans because they get to act on the AI’s output. The titans bought the same AI and surrounded it with a process that was designed to manage human campaign managers. The AI is faster than the process. The process wins, every cycle, by design.

Do this now

If you run marketing in a multi-brand or multi-region organization and your AI vendors are reporting impressive pilots while your P&L shows basically nothing, schedule the three diagnostics in the next two weeks.

Run the time-to-action trace on one campaign. Count the people, count the days. Compare to the cadence at which the AI produces new recommendations. If the recommendation arrives stale, the AI is not the problem.

Audit the approval chain by value, not by tradition. Each gate has to demonstrate a decision it changed in the last quarter. Gates that cannot are candidates for removal. This is uncomfortable because it surfaces work that exists to protect against accidents that have not happened in years.

Assign ownership of AI realization to a single person with budget authority. Not the CMO who bought the tool. The operator who runs the marketing P&L. The realization stops being invisible the moment one name is responsible for it.

The challenger brands did not win because their AI was better. They won because their org chart did not eat the value. Enterprise marketing teams have the same AI available. The next step is not another vendor evaluation. It is an honest look at why the capability they already bought cannot produce a return inside the structure they already have.

---

This analysis synthesizes Why AI Might Do More for Challenger Brands (Mike Shields, Next in Media, May 2026).

Victorino Group helps marketing organizations diagnose where org complexity blocks AI realization and design governance that enables rather than gates. Let’s talk.