Every major ad platform just launched an AI agent.
Yahoo DSP embedded agentic capabilities for automated audience discovery and campaign activation. Google launched Ads Advisor and Analytics Advisor for conversational optimization. NBCUniversal rolled out systems that autonomously orchestrate planning across linear TV and streaming. IAS introduced an agent that optimizes brand safety while keeping humans in the loop.
The message from every platform: the agent era is here. Hand over the keys.
The message from marketers: not so fast.
New data from eMarketer reveals a trust gap that every platform seems to be ignoring.
The numbers behind the skepticism
- Only 55% of marketers trust agentic AI to plan and execute tasks autonomously
- 1 in 5 marketers actively distrust fully autonomous systems
- Only 39% of marketing and sales professionals are confident that AI will drive revenue
- 46% remain ambivalent about AI's impact on their results
- 54% fear losing creativity and human touch when automation takes over
Let that sink in. More than half of marketers are scared that the thing their platforms are pushing will erode the core of what makes their marketing work.
This isn't irrational fear. It's a reasonable response to a real problem.
Why the trust gap is rational, not ignorant
Platforms have a history of overpromising
Programmatic advertising was supposed to eliminate wasted spend and maximize ROI automatically. Then brand safety crises happened. Attribution fraud happened. Automated bidding drove up CPMs in ways that benefited platforms more than advertisers.
Marketers have been burned before by automation that optimized for the wrong things. When platforms say "trust the agent," the historical memory of "trust the algorithm" kicks in. The skepticism is earned.
Agentic AI has a black box problem
When an AI agent builds your campaign, selects your audience, writes your copy, and sets your bid strategy — and the campaign underperforms — you can't easily debug why. Traditional campaign management is opaque enough. Agentic AI adds another layer of inscrutability.
Marketers aren't just worried about results. They're worried about accountability. When something goes wrong, who's responsible? The agent? The platform? The marketer who clicked "approve"?
The creativity fear is substantive
The 54% who fear losing creative and human touch aren't being sentimental. They're identifying a real risk.
AI agents are excellent at optimization. They're genuinely good at finding patterns in data and exploiting them. What they struggle with is the creative leap — the unexpected angle, the counterintuitive campaign, the brand story that doesn't optimize for any current metric but builds something that lasts.
The best marketing often breaks the pattern. Agents, by definition, optimize for the pattern.
Control is not the same as oversight
Most platforms claim their agentic features keep humans "in the loop." But there's a difference between rubber-stamping an agent's decisions and actually exercising meaningful oversight.
If an agent builds a campaign, and you're shown a summary with an "Approve" button — are you really in control? Or are you just a compliance checkpoint in an otherwise autonomous process?
Marketers can feel the difference. And they're not comfortable being a checkbox.
The gap isn't a trust problem. It's a control design problem.
Here's the reframe: marketers don't distrust AI agents because they're anti-technology. They distrust them because current implementations don't give them meaningful control.
The platforms that will win marketer trust are the ones that redesign the human role from "approver" to "director" — where the agent proposes, explains, and executes, but the human meaningfully shapes the strategy, constraints, and creative direction.
That's a different product than what most platforms are building.
What marketing teams should actually do right now
1. Adopt agents for the right tasks
There are agentic use cases where the trust gap doesn't matter much:
- Budget pacing and bid adjustments — optimization against clear performance signals
- Audience expansion testing — systematic variation within defined guardrails
- Reporting and anomaly detection — pattern recognition across performance data
- Routine content variations — generating 15 subject line options from a brief
There are use cases where it matters enormously:
- Brand voice and messaging — agents can match tone but miss nuance
- New market positioning — agents optimize for what has worked, not what could work
- Creative concepting — the unexpected angle requires human intuition
Don't deploy agents everywhere. Deploy them where they're genuinely better, and keep humans central where human judgment is the actual differentiator.
2. Demand explainability before you demand results
When evaluating agentic features, ask: can I see why the agent made each decision? Not a high-level summary — specific reasoning for specific choices.
If a platform can't explain why its agent selected a particular audience segment, bid strategy, or creative format, you're flying blind. And blind optimization is not a strategy.
3. Build your own performance baseline first
Don't adopt agentic AI into a campaign environment you don't understand well. Establish your own benchmarks — CPA, ROAS, conversion rates by segment — before you hand control to an agent.
Agents optimize against signals. If you don't know what "good" looks like for your campaigns, you can't evaluate whether the agent is actually improving things.
4. Treat the trust gap as a competitive advantage
Most of your competitors are either adopting agentic AI uncritically or avoiding it entirely. Neither extreme is winning.
The marketing teams that earn the next few years will be the ones that adopt agents selectively, maintain genuine strategic control, and use the efficiency gains to invest more in the genuinely human elements of their marketing — creative, strategy, brand.
That's a differentiated approach. And differentiation, as always, is what marketing is actually for.
The bottom line
The AI trust gap isn't a sign that marketers are behind. It's a sign that they're paying attention.
When platforms tell you to trust the agent, the right response is: show me the controls. Show me where I'm actually making decisions, not just approving them. Show me how this makes my marketing more human, not less.
The platforms that can answer those questions honestly will earn adoption. The ones that can't should expect skepticism to persist — and deserve it.
Need help building an AI-assisted marketing system that keeps your team genuinely in control? Let's talk →