Towards an Agentic Operating System for Physical Media

Google's Performance Max does not ask a media planner which inventory to buy. An advertiser inputs a goal and a budget, and AI determines allocation across Search, Display, YouTube, Shopping, and Maps — in real time, continuously, without human intervention at the execution layer. Meta's Advantage+ operates the same way: audience targeting, creative selection, and placement are all handled by the system. The planner sets the objective. The machine does everything else.

Mark Zuckerberg has been explicit about where this goes. He has stated publicly that all Meta advertising will be AI-planned and AI-executed. Not eventually. By the end of next year.

This is not a vision statement from a tech conference. It is the current operating model of the two largest advertising platforms on the planet, and it is accelerating. The question this raises for Out-of-Home and physical media is not "when will AI come to our channel?" AI is already planning and buying campaigns. The question is whether our industry will be connectable to those systems — or whether physical media remains invisible to them.

Two directions - Which one will you choose?

When an AI agent plans a campaign, it does not browse a PDF rate card or wait for an email from a sales rep. It queries structured inventory data, evaluates audience signals against campaign objectives, and activates supply through the appropriate demand channels — all in a compressed timeframe that assumes connected infrastructure is already in place. Inventory that is not structured, not measurable, and not connected to programmatic demand does not get priced lower. It does not appear in the plan.

This is the real divide forming in our industry — not between large and small operators, not between premium and standard inventory, but between media owners whose technology creates a machine-readable surface area and those whose does not.

On one side: operators with unified inventory management, dynamic ad serving, integrated audience data, open programmatic connectivity, and outcome measurement infrastructure. Their screens are visible to agentic buyers. Their inventory can be found, priced, activated, and evaluated automatically. They compete directly for digital budgets because they operate on the same terms.

On the other side: operators running disconnected platforms — a CMS here, a booking tool there, measurement living in a separate system or not at all. Their inventory may be premium by every physical metric — location, footfall, dwell time — but it is invisible to the automated workflows increasingly driving allocation decisions. Manual execution at scale is not a viable operating model when the market is moving toward automation.

Not if, but when

Eighteen months ago, building an agentic advertising platform was theoretical. Today, it is production-ready. The AI models, orchestration frameworks, and cloud infrastructure required to automate the full campaign lifecycle — plan, activate, serve, measure, monetise — all exist. The window to build connected infrastructure and to own that position is open right now. But windows in platform shifts do not stay open.

Early movers in infrastructure shifts tend to become the standard. When programmatic transformed digital buying, the publishers who connected earliest to DSP demand and adopted open measurement standards captured a disproportionate share of automated spend. Those who waited negotiated from weakness. The same dynamic is now playing out in physical media, and the pace of AI adoption suggests this cycle will move faster than the programmatic transition did.

Out-of-Home as a category is growing — more screens, more cities, more inventory entering the market. But growth without connected infrastructure does not automatically translate into incremental revenue. It creates more supply competing for budgets that are increasingly being routed through automated channels. That is not an opportunity. It is a yield problem.

What the Agentic OS actually means

The right framing is not digitisation. Most operators have already made that investment. The more useful question is whether the infrastructure functions as a system that AI can work within — a modular layer connecting every stage of the campaign lifecycle into a continuously learning loop.

A brief enters the system. AI recommends screens, audiences, and budget based on live data. Buying executes automatically through programmatic channels. Delivery optimises in real time against contextual signals — audience movement, dwell patterns, time of day, local events. Measurement closes the loop with outcome data, not just impressions. Every campaign makes the system smarter, and because that intelligence compounds, the operator who builds it first builds a moat that becomes harder to replicate with time.

This is not hypothetical. Jeki Japan demonstrated it by automating tens of thousands of transit screens into a single addressable marketplace. Publicis West Africa applied the same logic across 22 fragmented markets, turning what would have required separate local negotiations into one coordinated inventory pool. In both cases, the underlying driver was not the screens. It was the infrastructure connecting them.

For media owners, the implication is direct: inventory moves from being a product that sales reps sell to a data asset the market can discover, evaluate, and transact with automatically. Screens previously invisible to programmatic demand become part of plans digital buyers are already running. Inventory that was evaluated purely on location and visibility can now be evaluated on the outcomes it drives — which is how performance budgets are allocated.

This infrastructure does not require displacing existing systems. It requires connecting them. The gap is rarely the investment already made. It is the connective layer between those investments and the demand channels where buyers now operate.

The window is open, for now

The operators who will hold and grow their share of advertising budgets over the next five years are not necessarily the ones with the largest networks. They are the ones whose inventory can participate in agentic workflows — found automatically, bought programmatically, measured on outcomes, and continuously optimised. Nobody has yet unified the full stack for physical media into an intelligent operating system. That category is still being formed, and it belongs to the operators who move first.

The question for every media owner is simple: when an AI agent allocates a campaign budget next quarter, will your inventory be in the plan?

About the Author

Mehul Mandalia is Co-Founder of Moving Walls, a connected media company building autonomous planning and measurement infrastructure for Out-of-Home and physical media.