TV advertising was built for a simpler world twenty years ago: a dozen sellers, a few currencies, 30 second spots and human driven workflows. With broadband, mobile phones, smart TVs with streaming and social/vertical video, the operating system of TV advertising now runs on continuous, high-frequency data—delivery signals, identity hints, content metadata, supply paths, outcomes—arriving faster than any team can interpret manually. Modern video is, at its core, a big data opportunity where algorithms aren’t an innovation layer, they’re the steering wheel.
Ten years ago, “AI in advertising” mostly meant better prediction: smarter targeting, lookalikes, incremental lifts in bidding and measurement. Useful, yes—transformative, not quite. What’s arriving now is agentic: systems that can plan, activate, monitor, verify, and reconcile spend with far less human glue. And that’s a very different kind of disruption, because it touches the highest-friction part of the industry: how decisions get made and dollars get committed.
If you think the agentic era will be won by whoever has the best prompt, you’re going to miss the plot. The winners will be the companies—and the industries—that build the rails: standards, permissions, metadata, contracts, APIs, and audit trails. Models get headlines. Infrastructure builds trust and generates compounding returns.
From optimization to orchestration (and why standards suddenly matter)
We’ve already lived through one automation revolution. Programmatic turned unsold ad inventory buying into software; machine learning turned into an optimization engine. I saw it up close at Right Media in 2004 when we launched dCPM and later the Right Media Exchange (RMX). The lesson wasn’t inevitability—it was prerequisites: automation scales only after standards and systems make transactions consistent, interoperable, and trustworthy at low cost. That last requirement is why the industry is so focused on streamlining the programmatic stack as TV becomes more digital and direct IOs and programmatic lines blur.
That’s why efforts like AdCP—and the hard, unglamorous work from IAB Tech Lab—matter more than people realize. Agents will need consistent ways to request inventory, declare constraints, pass instructions, and receive outcomes. Without shared semantics and predictable interfaces, “agentic advertising” becomes a bunch of impressive demos stapled to a fragile supply chain.
We can also see the market separating into roles. Some companies will build domain agents—creative versioning, supply-path decisions, budget pacing, anomaly detection, forecasting. Others will build orchestration layers that coordinate many systems. And the platforms will embed agents directly where decisions happen, because control over the “moment of commit” is where the economics live.
It’s also worth saying out loud: none of this is magic. It’s capex. The largest companies have spent staggering amounts on compute, networking, and AI infrastructure to turn frontier models into something you can actually build a business on. That spending is why capabilities we used to daydream about—systems that can absorb context, reason over constraints, generate alternatives, and take multi-step action—are now table stakes.
In an agentic world, the scarce asset right now is governance
As agents get better, faster, cheaper, governance becomes more critical and valuable. When software can commit dollars in milliseconds, “trust me” and manual intrusion stops working. Enterprises must have permissioning, approvals, contract enforcement, clear liability, and an auditable chain from intent → execution → invoice.
Automation increases the premium on proof. When decisions are machine-made, buyers need to know what actually ran, where it ran, how often it ran, and whether it met quality and suitability requirements. That’s where independent sources of truth matter. The durable platforms will connect authority and evidence into a closed loop: plan and approve, activate, validate, measure, and reconcile—with accountability at each step. That loop is what turns AI from insight generation into core operations.
Agentic ROI hinges on systems integration
Advertising is approaching a trillion dollars in global spend. If we want CFOs and CMOs to treat advertising as a repeatable growth lever—not a perpetual reconciliation exercise, or worse, optional for the long-term health of a company—we need the foundational infrastructure to match the ambition. Standards like AdCP and IAB Tech Lab’s work support sturdy rails. AI capex is the fuel. The control plane keeps the trains on the tracks.
Agents won’t create value by optimizing one task in isolation, and moat isn’t the UI. It’s the authority, governance, and audit trail that interoperate and create value by orchestrating dozens of tasks under constraints—across planning, audiences, creative, activation, measurement, and finance. The control plane is architectural: it connects those steps and reconciles what was intended, what executed, and what got paid. That’s how you get automation that’s continuous, accountable, and auditable, not just fast.
In a world where many AI systems are aligned to a single walled garden, buyers will insist on a control layer that works across partners and channels without hidden incentives. Neutrality becomes a stronger feature. Supporting third-party agents while building native ones is the right posture: let innovation happen at the edges, but keep execution grounded in governed workflows. Smaller advertisers may not require this; the largest mandate this.
We are at a moment of profound change in our industry. And while there is likely a trough of disillusionment coming based on the timing of expectation versus the reality of how quickly AI is adopted and driving value for companies, the impact of AI on advertising is largest than anything we’ve seen before. We have a lot of good work to do together.
