Apostra Wants AI Agents to Negotiate the Media Plan

Scope3 became Apostra on September 23, and the new name marks a change in what the company sells. If your B2B team buys a newsletter or podcast sponsorship, the pitch is for your AI agent to ask that seller for a tailored offer, compare it with other offers, and bring the proposed buy back for your approval. The launch announcement says the offer spans digital, social, television, audio, outdoor, and print.

That is a different job from asking AI to adjust bids on the ads your team already buys. Apostra is trying to change what gets considered for the plan in the first place: a podcast sponsorship, a local billboard, or a custom publisher package that would never appear in a standard display auction.

Co-founder and chief operating officer Anne Coghlan put the technical hurdle plainly in a September 23 essay:

“An agent doesn’t need every media owner to have an API.”

Instead, she wrote, it needs a common language for finding and talking to those sellers. That language is AdCP, the Ad Context Protocol, an open standard for agents to discover, plan, buy, sell, and measure media. Apostra is the commercial place where the conversations can happen.

The brief becomes the buying interface

A marketer gives the buying agent an objective, audience, budget, and rules. A media company lists its properties, formats, audiences, packages, and prices. Its sales agent can answer the brief with a proposal that includes a package or sponsorship, not just a row of available impressions. Apostra’s advertiser page says the buyer can ask one seller or several, compare their terms in the same structure, and turn a chosen proposal into a draft media buy. The seller and terms carry through to that draft.

The human still sets the boundaries. Apostra says advertisers can specify exclusions and approval thresholds, while media companies can set pricing and decide which campaigns need sign-off. Those controls matter because a buying agent with permission to spend has a different failure mode from an agent that only makes recommendations. A marketing team should know whose rules get checked before a proposal becomes a booking, and who can stop it when the offer changes.

This is the step beyond the agentic auction data we covered in August. That story measured what happened when software picked among digital ad impressions. Here the agent is asking a seller what else it could offer. It is also a different route from Google’s Buyer Direct, which lets buyers book guaranteed publisher inventory inside Google’s own ad server. Apostra’s pitch is that the brief can travel across sellers and media types instead of ending at one platform’s inventory.

Co-founder and chief executive officer Brian O’Kelley describes the model as an ad gateway. In a September 24 essay, he wrote:

“Our goal is to have every single advertising surface, every media company, and every creator available through the gateway”

The ambition is wider than the current proof. O’Kelley says the platform launched with 40 media partners, but the company hasn’t published a partner-by-partner inventory list or independent results for its buys. Apostra says it takes 0% of media spend, with optional paid intelligence still in beta and pricing to come. That’s a pricing claim from the seller of the service, not a verified total-cost comparison against an agency or demand-side platform. A media buyer still needs to ask who charges for the proposal, the execution, the ad server, and the measurement.

More choice creates a harder comparison

The attraction for a brand is real. A single brief could surface offers that its team would not have had time to request one by one. A seller could describe a valuable audience or sponsorship in its own terms rather than flattening it into a standard impression. Coghlan argues in her essay that an agent can read a rate card, email a sales team, or fill an insertion order when a seller has no direct technical connection. That is a plausible way to widen the market without waiting for every local media company to build new software.

It also makes comparison harder. A podcast host read, a TV spot, and a digital display package have different evidence of reach and different paths to a sale. A neat proposal table won’t make those outcomes interchangeable. A team that treats the common format as a common measure could buy the cheapest-looking package rather than the one most likely to meet the objective.

Kenneth Rona, chief AI officer at JWX, was skeptical about the initial launch material. In a September 24 newsletter post, he wrote:

“I did not force a story on it as there is not much to tell.”

His point is a useful check on the pitch. A working place to request and compare offers is one step; evidence that the offers beat a human buyer’s plan is another. The launch didn’t supply that evidence.

For the next media plan, the useful test is small. Give one agent a brief for a defined audience and budget, ask for offers from two kinds of sellers, and compare its plan with the one your buyer would have built. Record every fee, the approval point, and the measurement each seller can actually deliver. Apostra has made the brief portable. Whether that produces better media buying is a result marketers still have to measure.