Show me the ROAS: What Jerry Maguire can teach us about agentic media buying

Every marketer secretly loves this movie scene: Jerry Maguire, fresh off his meltdown, screaming into the phone for someone to just show him the money. Dramatic. A little unhinged. Also a perfect parallel for every board meeting about media performance since digital advertising began. The ad industry has a new kind of agent walking the halls, and it’s not Bob Sugar. It’s algorithmic, and it’s forcing agencies to ask the same question Jerry asked himself in the mirror: what do we do here, and who are we doing it for?

Enter the agentic agent

Agentic media buying means AI systems that don’t just recommend a bid or a budget — they act. Google’s Performance Max and Smart Bidding, Meta’s Advantage+ and Andromeda and a new wave of proprietary agency tools are making thousands of micro-decisions a second: budget shifts, creative optimizations, audience expansion and bid adjustments. This is beyond “AI-powered” or “automated.” The system perceives a goal, decides, acts and adjusts based on the result. Earlier generations of automation still needed humans to write the rules. Agentic systems set their own path. Pros and cons aside, agentic media buying has officially entered the chat.

What sports agents actually got right

A great sports agent was never valuable for throwing the touchdown or hitting the jump shot. They were valuable because they saw what the athlete couldn’t from the field: market timing, leverage, positioning, and the gap between a good contract and a career-defining one. The athlete performed. The agent turned that performance into outcomes. Agentic AI in media buying is the athlete. It executes at a scale and speed no human buyer can match. But it doesn’t understand the client’s business the way a strategist does. It doesn’t know that Q4 margin pressure means efficiency beats volume this quarter, that a category is about to get disrupted by a competitor spending more, or that the CMO just got a new boss with different priorities. It optimizes toward the objective it’s given; it doesn’t set the objective. That’s the value the modern agency brings: not running the auction, but representing the client in it. The same way Jerry Maguire never played a down, but made sure Rod Tidwell got paid what he was worth.

Where the value actually sits today

The efficiency gains are real. Agentic systems find incremental performance humans can’t find manually, at a pace humans can’t match. But the black box problem is real too. “The algorithm decided” is not a strategy, and it’s not an answer a CFO wants to hear in a quarterly business review. The agencies pulling ahead right now aren’t the ones resisting agentic tools, or the ones blindly trusting them. They’re the ones building the translation layer between what the machine is doing and what the business needs to know. That’s not a small skill. It’s the whole skill.

Agencies as the front office

If agentic AI is the athlete, the agency’s future looks less like a media-buying desk and more like a front office. Develop the roster strategy, manage the negotiation, read the market and make the judgment calls no algorithm is positioned to make, including when to override it. That means the paid media buyer of the near future spends less time in the interface and more time doing what a good agent has always done: understanding what the client needs to win, translating machine output into business language and knowing when the “optimal” algorithmic decision is the wrong one for the brand, the quarter or the moment. The agencies that get this right won’t be competing with the algorithm. They’ll be managing it, the way a great agent manages a roster of talent.

“Help us help you”

Agentic media buying frees agencies from an ever-increasing load of manual execution. That time goes back into problem-solving, context and partnership —the reciprocity between data and strategy that the modern marketing machine runs on. The result: more trust, built through strategic thinking, delivered through transparency.

The machines can run the plays, but the agencies still need to call them.