German publisher sales house BCN and PubMatic are removing the manual plumbing from private marketplace deals. Their new Signals Agent lets an advertiser type a plain-language audience description, then hands the matching BCN segments to PubMatic’s AI agents to create and activate a pre-targeted PMP deal.
From brief to deal without the spreadsheet relay
Instead of assembling targeting across disconnected tools, buyers describe who they want to reach. BCN’s example: “I want to reach people with a high income who are currently interested in electromobility and in buying a premium vehicle.” The agent matches that brief semantically against more than 1,500 signals: 900-plus audience segments and 600-plus contextual categories from 80 million consented profiles across 50-plus premium domains.
BCN says the system ranks relevant segments, returns reach estimates and prices, then communicates directly with PubMatic’s agents via MCP and the Ad Context Protocol. PubMatic builds a pre-targeted private marketplace deal and activates it on the exchange. No manual data hand-off between systems, BCN claims.
Why this matters for performance teams
This is not a cosmetic AI layer. It moves campaign setup from human-to-human negotiation to agent-to-agent deal creation. For advertisers, that compresses a traditionally slow workflow into a few minutes and shifts targeting decisions to the sell side, where BCN’s data defines the audience and the price.
The deal still needs a buyer’s DSP to bid. BCN did not disclose which DSPs are enabled, how deal IDs reach the buyer, whether agencies must accept the deal, or what floor prices apply. Those mechanics will decide whether this becomes a genuine buying shortcut or another curated deal that needs manual babysitting.
What to ask before you test
- Identity basis: BCN says more than half of signals rely on deterministic IDs, but it doesn’t name the identifier framework. Ask exactly what “profile” means and whether it’s a person, browser, or device.
- Governance: BCN says selected signals are checked against policies before activation, but doesn’t specify whose policies apply or whether a human approves the deal. Request the approval trail.
- Commercials: Are returned prices a CPM floor, data fee, or fixed rate? Agency and client teams need that split before comparing CPMs.
The bigger agentic shift
PubMatic has been building this story all year: AgenticOS launched in January, and the company reported more than 80 fully autonomous campaigns and 4,000-plus AI-powered deals since then. Independent DataBeat data from May 2026 showed agentic demand appearing in 86% fewer auctions and clearing at $6.13 CPM versus $6.95. That is closer to TV-style curated buying than broad open-auction bidding.
For D2C growth teams and agencies, that is the practical takeaway: agentic deals will likely be fewer, more curated, and priced around sell-side packaging. Test one PMP before scaling, compare total effective CPM, and track whether faster setup actually produces better ROAS or just faster setup.
The implementation is bilateral today: one German sales house, one exchange, one pair of agents. The promise of interoperability is real, but it stays a beta until buyer agents from other platforms can query BCN directly and BCN signals can activate outside PubMatic. Carsten Sander, BCN’s managing director, framed the product as reducing a laborious process to “a simple question: which target group do I want to reach?”
Source: PPC Land
