OOH Wins Agent Budgets By Publishing Data A Machine Can Query

How bright a billboard is, how many people pass it, and why they're there all decide how well it works. Much of that sits in databases already, split across suppliers in formats that don't line up, with no single place an AI agent can query. A digital ad slot arrives with the same kind of detail attached and readable. An agent that can't look up a billboard won't buy one. Out-of-home still reaches the people other channels are losing access to, and the work now is pulling those facts into one place an agent can use.

Shawn Spooner is Global Chief Technology Officer at billups, where he has spent more than a decade building the analytics and measurement platform behind the company's managed services work. Spooner joined as Chief Scientist in 2014 and built the research team, including the first model for direct return-on-investment attribution in the category. He holds patents across artificial intelligence, computer sensing and media measurement, and co-founded healthcare machine learning companies before that. His view is that agents route around the companies they can't transact with.

"My read on agents in general is that they'll go through hurdles as long as they can get the answer themselves. If they hit a dead end of having to go to a person, they'll start working around that person. It's not malicious behavior, they're just goal oriented," says Spooner. A marketer tells an agent to raise brand awareness with a set budget, and out-of-home may surface in what it finds. Most out-of-home inventory is still sold person to person. An agent buys the small share available programmatically and leaves the rest. Making the rest reachable means describing it, pricing it and proving it in terms a machine can check.

Signals without a schema

Online, an ad system knows what someone searched for and which page they landed on, and that's most of what it needs to guess at intent. The same crowd passes a screen whether they're commuting, shopping, or killing time. The placement performs differently in each case, but nothing in the current data tells an agent which is correct. "The context of the physical world is really not that discoverable to agents, and out-of-home lives in that hard to discover space," Spooner says. "It provides a lot of opportunities for building systems that make that space more legible."

Spooner describes an agent weighing a site with a list of who walks past it and no read on what the environment does to their attention. Two placements reaching identical demographics can perform very differently, and the difference sits in conditions no current data feed reports. "You can buy it on the audience alone, but you're missing a whole signal on what are they probably there for," Spooner notes. "There's not a good single solution right now that surfaces that in a legible way."

The rails underneath

Before an agent can buy anything, it needs a way to pay that a finance team will accept. Mastercard and Stripe are both building that layer for agentic commerce, issuing single-use cards an agent can spend against and settling transactions in stablecoins. The card expires with the transaction, and the risk stays contained. "That's a huge investment from companies that typically are very good at building boring things," Spooner explains. "The fact that they're leaning that heavily into this tells me they actually think there's a lot of value here."

An agent acting on a buyer's behalf has to leave a trail. Spooner wants receipts for every step between the objective and the placement, in a form a person can audit after the fact. A campaign that worked and one that stumbled into the same numbers look identical on a results screen. "I've got to understand why it reached the decision that it reached, what data it used, how was that data grounded," Spooner says. "Can I work backwards and figure out, is this a good chain of events that led to this outcome, or did I just get lucky?"

Proof in the open

An agent told to check whether out-of-home works will go looking for evidence. Spooner expects it to hit a paywall on the independent research within a few searches. What stays free comes with a question attached about who paid for it. "It's going to find that it's all been generated by out-of-home suppliers," Spooner notes. "That's an axis on trust that isn't as strong as one that was done completely independently."

Spooner's proposal is a shared pool of older studies, anonymized and stripped of client detail, hosted somewhere an agent can query. Studies a year or two old have already served the clients who paid for them. He puts the hosting cost near zero and the legal terms as the hard part. "If there's a hundred studies or a thousand studies that all point the same direction, that's probably the right truth," Spooner explains. "The broader pool that we do, the better answers come back."

Spooner watches the effect on his own inbox, where an agent screens what arrives. Direct emails that used to reach him ten times a day now reach him about once a week. Out-of-home holds its audience through all of it, and that is the advantage he wants the industry to act on. "For now I can't ad block out-of-home," Spooner concludes. "We're going to have to make the space more legible, more transactable, more measurable, and those are the things that are going to make that transition happen."