Foreword
Why we ran this study.
The word “agentic” now covers everything from a copy assistant to a system that changes bids without being asked. Those are not the same thing, and the difference decides whether AI is a productivity tool or a change to how marketing operates. This study drew a hard line between them and counted only AI that takes action inside marketing tools.
What it found is a gap between how widely AI has been adopted and how little of it runs at scale. Almost every team is piloting or deploying, very few have expanded what already works, and almost none let AI act without a person in the loop. What holds it there is not the technology but the operating model around it: data that is not ready, budgets AI has to borrow, controls that rest on a person signing off, and success still counted in hours saved rather than business results.
What follows is not a forecast. It is a description of where 100 senior marketers stand today, what they say has to be true before they go further, and what the minority already furthest ahead do differently.
Almost everyone is piloting or deploying agentic AI. Only 1 in 8 has scaled it. What is holding it back from scaling?
The constraint is not appetite, and it is not the technology. It is the operating model.
Executive Summary
What we found.
Among marketers already working with AI, it is a priority almost everywhere and at scale almost nowhere. Nearly all of them call it a priority, but only 12% have reached scale and just 5% let AI act on its own inside guardrails. The constraint is not appetite, and it is not the technology. It is the operating model: AI is judged on time saved rather than business results, funded from budgets it borrows, controlled by human sign-off, and running on data that only 2% call ready. Marketers know what would change that, and it is a condition their current operating model cannot meet: 80% say they need demonstrated incremental lift before expanding beyond a pilot, while only 5% use causal methods to judge what to scale. A minority already work differently.
- 12%have reached scale
- 5%let AI act automatically within guardrails
- 2%call their data consistently ready for AI
- 80%need proof of incremental lift to scale past pilot
Key Findings
The four findings.
The State of Play
AI is on every agenda and almost nowhere at scale. Teams are piloting and deploying, barely one in eight is expanding what already works, and AI that acts on its own is rare. Ambition is not the constraint.Where AI Is Working Today
AI has landed where the work is high-volume and lower-risk: paid media and creative. It targets, drafts and reports. It does not decide, it rarely touches the budget, and only 8% of workflows run end to end.The Operating Model
AI is judged on hours saved rather than business results, funded from budgets it borrows, and controlled by people signing things off. Underneath it, the data is not ready and there is no settled way to decide, govern or approve.The Year Ahead
Marketers expect more of the same, faster: efficiency and effectiveness ahead of anything new. What they say would unlock the next stage is proof of incremental lift; what they expect to hold them back is their own organization, not the technology.
Further Ahead
The marketers who are further ahead.
One group behaves differently, and it is not defined by company size or budget. It is defined by proximity to the work: 31% describe their own knowledge of how AI is used in their organization as detailed and hands-on, rather than a good working understanding.
They expand on proof, not permission
They want reliability demonstrated before they scale something (74%, against 44% of everyone else), and they need executive sponsorship to do it less often than the rest.
They measure in money
They judge AI on ROAS and cost per acquisition, in a market where 74% still count time saved.
They have built the controls others lack
The controls that let AI act without someone watching, kill switches and audit logs with a named owner, are several times more common inside this group than outside it.
Leading from the Inside
Hands-on marketers run AI on evidence, not permission.
31% of marketers have detailed, hands-on AI knowledge, and they operate differently.
Segmented on self-reported knowledge of how AI is used in the organization, not on maturity or company profile. The group is also further along on maturity and on data readiness, but it was selected for proximity to the work rather than for being advanced, so the direction of that relationship cannot be established here. Hands-on n=31, everyone else n=64; the segment is small, so these differences are directional and are not significance-tested.
Methodology
What this study asked.
The definition every respondent read before answering. Everything counted in this report is AI that acts; AI that only drafts or recommends is measured separately and reported as such.
What it set out to answer
- How far along is agentic AI in marketing, really?
- Where is AI working today, and where is it heading next?
- What is holding it back from scaling?
Base: 100 senior marketers. To qualify, respondents had to own or influence AI strategy, tooling or execution in marketing, and confirm first-hand knowledge of the platforms used, the workflows automated and how outcomes are measured. Because of that screen, the study describes how AI is used by marketers already using it; it does not measure how widely marketing as a whole has adopted AI. 88% are Director level or above, 63% set AI strategy or budget, and 95% are headquartered in North America. Full profile in the Sample & Respondent Profile.
Chapter 1
The State of Play
How seriously marketing takes AI today, how mature teams are, and how AI operates in the work.
The full report
Continue reading
6 chapters · 37 interactive exhibits
- 01 The State of Play
- 02 Where AI Is Working Today
- 03 The Operating Model
- 04 The Year Ahead
- 05 Leading from the Inside
- 06 Conclusions
Not an MMA Member? Please fill out the form below to read this report and learn more about other useful resources.
MMA Members, log in to read now.
Thank you for your submission.
State of Agentic AI in Marketing 6 chapters · 37 interactive exhibits
Continue reading
