Most marketing teams use AI and get no business value from it. When something underperforms, the reflex is to add another tool. That is the wrong move: the group pulling ahead uses the same stack you do. The difference is that they can show where AI produces value and where it does not, and you cannot know that without measuring it. Value leaks through four gaps: evidence of value, data readiness, team adoption and governance. Almost nobody has checked which of the four is costing them value.
- AI adoption in marketing is close to universal, but only a small fraction gets business value from it.
- The fraction pulling ahead runs the same stack as everyone else. The difference is system, not software.
- Value leaks through four failure points: evidence of value, data readiness, team adoption and governance.
- Most dashboards measure AI activity (pieces produced, hours saved), not AI value (better decisions, revenue influenced).
- The first move is to diagnose which of the four failure points is costing you value, before adding any new tool.
The wrong reflex: add another tool
A single figure keeps appearing in the 2026 reports, and it cannot be ignored.
AI adoption in marketing is close to universal. The share that has genuinely embedded it into how they work is a small fraction. The gap between using AI and getting value from it is wide, and it is growing. The comfortable explanation is that the fraction pulling ahead has better tools. It does not. It runs, for the most part, the same stack as everyone else. The difference is one of system, not software. And the first symptom that the system is failing is that the team cannot answer a simple question: where, exactly, is AI producing value?
When AI does not deliver the expected results, the near-automatic reaction is to look for the next tool. Another model, another agent, another platform that this time will deliver. It is an understandable reflex. It is also the one that keeps a team going in circles. In an earlier post I argued that AI is mostly applied to the execution layer, producing content, where it creates efficiency rather than advantage. That was a question of where: which layer of the system you point AI at. This post tackles the next problem. Even when you apply it in the right layer, the value can still leak away, because nobody has measured what actually produces it.
Adding one more tool without knowing where the value is lost is like swapping the car when what you have is the wrong map. The car was never the problem.
Feel pressure to adopt AI, and reach for the next tool when it underperforms. The assumption: a better tool would fix it.
Have genuinely embedded AI into their workflows. The group that gets value runs the same stack. Measure, do not buy.
The group pulling ahead with AI uses the same stack you do. The difference is that they know where it produces value. You have not measured it. — Reyes Brusola
Value does not leak in one place. It leaks through four gaps.
When a team cannot turn AI into business, the leak is not random. It runs through four failure points.
Most teams have no idea which one is costing them value because they have never looked.
Evidence of value. This is the question the board asks first and the one most teams cannot answer. Almost everyone can describe which task AI sped up. Very few can point to which business result it moved. Speed is easy to see and easy to confuse with value. If you cannot name a revenue line that AI influenced, you do not have a tool problem, you have an evidence problem. The 2026 data confirms it: 4 in 10 teams cannot prove marketing ROI across channels, and almost half say measuring it is their single biggest challenge. [1]
Data readiness. An agent acts on the data you give it. Weak data does not produce weak outputs. It produces confidently wrong ones, at speed. And here is the hole almost nobody looks at: half of marketing teams do not own their own data strategy. [1] As Supermetrics puts it in its 2026 report, you cannot build on a bad data foundation. Before building anything impressive with AI, that is the boring part almost nobody fixes.
Team adoption. AI that the team does not trust is quietly ignored. Adoption is a people problem before it is a technology problem. Introducing it badly burns trust that is hard to rebuild. The challenge leaders name is not the stack, it is how to bring AI into the team without people feeling their judgement is now surplus. Human judgement does not vanish with automation: it becomes the filter that decides whether what AI produces is useful or not. [2]
Governance. An agent acts autonomously, without direct supervision. Without governance, you only discover what the agent has done after the fact. The real danger is not the obvious error. It is the plausible-but-wrong output, produced at volume and waved through by default. Without a validation step, it becomes your published work before anyone reviews it.
Four failure points. Most teams leak in at least one, often several. And almost none know which, because measuring your own system is not what gets rewarded. What gets rewarded is adding the next tool.
We do measure. We have a dashboard full of AI metrics.
Most dashboards measure output, not value. Activity is not the same as better decisions.
Almost all of those dashboards measure output: pieces produced, hours saved, prompts run. None of those numbers tells you whether AI changed a decision, sharpened a segment or caught a pattern nobody had seen. Measuring AI activity is not measuring AI value. They are two different things, and confusing them is exactly how a team convinces itself it is doing fine while the value leaks away.
| The four gaps | Where value leaks | Symptom you are leaking there |
|---|---|---|
| Evidence of value | Speed gets mistaken for value | You can name the task AI sped up, not the revenue it moved |
| Data readiness | Agents act on weak, unowned data | Half the team does not own its own data strategy |
| Team adoption | AI the team does not trust is ignored | The tool is a side experiment, not part of the workflow |
| Governance | Plausible-but-wrong output ships by default | You discover what the agent has done only after the fact |
Why almost nobody measures it (and why that is the opportunity)
Measuring your own AI system is uncomfortable, so most teams never do it.
It forces you to admit that part of what was presented as progress was activity. It forces you to look at the boring foundations (the data, the governance) instead of the impressive automations. And there is no standard instrument for doing it, so most teams do not. That is where the advantage sits. The group that gets value from AI is not the one with the most agents. It is the one that treats AI as a system with evidence, data, adoption and governance, and keeps the four in balance as it scales. That starts with knowing which of the four gaps is costing you value today. It is not a question of abstract strategy. It is a diagnosis you can run in an afternoon.
Before you add another tool, audit your system
If you are not certain where AI is giving you value, that is exactly the starting point.
I designed a test that runs through the four gaps where value leaks and tells you which one applies to you. Twelve questions, eight minutes, each one with its source. It is not a pitch. By the end, you will have identified your weak point and the next step that matters.
Download the test: The Agentic Marketing Reality Check.
Frequently asked questions
Why most teams use AI but get no value from it
AI clusters in the execution layer, and value leaks through four gaps (evidence, data, adoption, governance) that almost nobody audits. The stack is not the differentiator. The ability to measure where AI produces value is.
Why having AI metrics is not enough
It depends what they measure. Most dashboards measure output (pieces produced, hours saved), not value (decisions changed, revenue influenced). Measuring AI activity is not the same as measuring its business value.
The first step to fix it
Diagnose which of the four gaps is costing you value before adding any new tool. Without that diagnosis, every extra tool is a blind bet.
Whether this applies to small companies or only large teams
It applies to any B2B team that already uses AI and cannot demonstrate the business value it produces. Size changes the scale of the problem, not its nature.
Sources: [1] Supermetrics, ‘The 2026 Marketing Data Report’, Supermetrics, 2026 (survey of 435 marketers across the US, the UK, Germany, Australia and Singapore). [2] Spencer Stuart, ‘The AI Reckoning: Why Marketers Think 2026 Is a Make-or-Break Year’, Spencer Stuart, December 2025. [3] Amit Kharche, ‘Why AI ROI Is Not a Metric Problem, It Is a Leadership Design Problem’, Medium, 2026.
Free · 8-minute audit
Find where your AI value is leaking.
The PDF arrives when you confirm your subscription. 12 questions, 8 minutes, the 4 value gaps.
Send me the checkEvery question with its source. One email a week. One-click unsubscribe.
