The automation paradox in advertising

This week brought a striking juxtaposition in ad tech. Ad Age reported that AI agents are now hallucinating ad prices and incorrectly booking media in ad servers. Meanwhile, Google rolled out new AI Max features and Meridian GeoX — a suite of measurement tools designed to push machine intelligence deeper into campaign operations.

More AI power. More AI mistakes. This isn't a contradiction. It's the reality of an industry in rapid transition.

The rush toward full automation

May 2026's ad tech landscape tells a clear story: advertising is being rebuilt around AI systems. Platforms are replacing manual campaign setup with machine-led logic. Bidding, audience expansion, creative generation — the space for human intervention is shrinking by design.

For small and mid-size marketing teams, this seems like a gift. You no longer need a dedicated media buying team to manage cross-platform campaigns. AI handles the complexity.

But here's the catch: when AI makes mistakes, who's watching?

AI agents have been caught fabricating ad pricing data and executing incorrect bookings. This isn't theoretical risk — it's happening now. When an AI system can autonomously move from planning to execution without human checkpoints, a single error can waste tens of thousands in ad spend.

Why human-in-the-loop isn't a step backward

The best-performing ad operations in 2026 aren't fully automated. They follow a human-in-the-loop (HITL) model:

1. AI analyzes and recommends — processing massive datasets, identifying patterns, generating optimization plans

2. Humans review and decide — evaluating whether recommendations make sense, guarding brand safety, confirming budget allocation

3. AI executes and monitors — carrying out approved changes with precision, tracking performance in real time

The critical design principle: AI proposes, humans approve.

This isn't about distrusting AI. It's about understanding its current limitations and building workflows that leverage its strengths while containing its failure modes.

Practical takeaways for marketing teams

Establish a review gate before execution

No AI tool should modify your ad account settings — bid adjustments, budget reallocation, audience expansion — without explicit human confirmation. Automation should accelerate the recommendation cycle, not bypass the approval step.

Use cross-platform AI insights, not siloed dashboards

Google's new Meridian GeoX improves cross-platform attribution, but each platform's native tools see only their own data. The real advantage comes from unifying data across Google, Meta, TikTok, LINE, and other channels to find optimal budget distribution across your entire media mix.

Talk to your data in natural language

Traditional dashboards require you to know what to look for. Modern AI tools let you ask directly: "Which audience segment had the highest ROAS last week?" or "What happens if I shift 20% of budget from Google to TikTok?" This interaction model dramatically lowers the barrier to data-driven decision making.

Prioritize creative optimization

As platform algorithms grow more sophisticated, creative quality is replacing bidding strategy as the primary performance variable. AI can analyze which creative elements — color schemes, copy tone, CTA placement — drive better engagement. But the final creative decisions still benefit from human brand intuition.

The opportunity for lean teams

This AI wave is actually favorable for smaller businesses. Cross-platform ad optimization used to require a full team: media buyers, data analysts, creative designers. Today, a lean team paired with the right AI tools can achieve operational efficiency that previously required 5-10 people.

The prerequisite: choose tools that keep humans in the decision loop, and resist the temptation of fully autonomous systems that promise to "handle everything."

In 2026, where AI agents still hallucinate, the smartest strategy is clear: let AI do what it's best at (speed, scale, pattern recognition) and let humans do what humans are best at (judgment, intuition, brand stewardship).

The bottom line

The winning teams in ad operations won't be the most automated. They'll be the most effectively collaborative — human intelligence and artificial intelligence working in a tight loop, each compensating for the other's blind spots.

Full autonomy is the destination. Human-in-the-loop is how we get there safely.