TL;DR: The term "agentic advertising" is now used by platforms with fundamentally different architectures. Apostra (formerly Scope3, rebranded September 2026) positions as a discovery and buying marketplace where agents can find and transact inventory. Large demand-side platforms use AI to assist human buyers within existing infrastructure. Neither model addresses agent-to-agent deal negotiation, which is a structurally distinct problem. Buyers evaluating agentic claims should ask three questions: who owns the deal record, whether the negotiation is auditable by both sides, and where the human approval gates are actually enforced.

The term "agentic" is now appearing across advertising technology, but the companies using it mean fundamentally different things. Apostra, the company formerly known as Scope3, rebranded in September 2026 around a single claim: "built for advertising when agents do more of the work." The Trade Desk positions around AI-powered omnichannel buying. And beneath both sits a third problem, largely unaddressed: how a deal negotiated between two agents is recorded, owned and governed when neither party is a human reading a screen. That is a structurally different problem than discovery or optimisation, and it is the one most likely to decide what "agentic" ends up meaning. For buyers and media owners evaluating what the word means for their operations, telling these three definitions apart is the most important thing to do before choosing a platform.

Why is the word "agentic" appearing in advertising now?

Advertising technology was designed for human operators. Demand-side platforms present bid management interfaces for traders. Supply-side platforms give yield managers controls over floor prices and deal terms. The measurement, verification and reconciliation systems sitting between them assume that a person reads the output and makes the final call.

Agents do not work this way. An agentic system, in the technical sense, reasons about objectives, takes actions and adjusts based on outcomes. When an agent represents a buyer or a seller, it needs to query inventory, negotiate terms and complete a transaction without a human reviewing each step. The existing advertising stack was not designed to support this. Different companies have diagnosed the resulting problem differently, and their diagnoses have produced different products.

What does Apostra mean by agentic advertising?

Apostra's current positioning, confirmed from the company's own homepage in September 2026, describes "one place for every kind of advertising, discoverable and buyable in one place" and "human control, agent scale." The framing is marketplace access: agents should be able to find and purchase inventory across digital, CTV, audio and out-of-home formats through a single interface, with humans retaining the authority to set the objectives and rules that agents act on.

This is a coherent and valuable proposition. An agent that cannot efficiently discover the right inventory across formats is a limited agent. Apostra's rebrand from a carbon-measurement identity to a broader advertising marketplace signals that the company believes this discoverability layer is where the structural opportunity now sits.

Apostra's positioning stops at discovery. The layer beyond it, how the negotiation between a buy-side agent and a sell-side agent is recorded, owned and governed, is a separate problem, and not one a discovery marketplace is built to solve.

How does AI-powered buying at scale differ from agentic negotiation?

The Trade Desk and comparable large demand-side platforms position AI as a speed and optimisation layer within an existing buyer-seller structure. AI processes signals faster than human traders, adjusts bids in real time and surfaces patterns that a person would miss. The Trade Desk's current public positioning describes an "omnichannel platform" with AI capabilities: better decisions at machine speed, but still decisions made within a platform that one party, the buyer, controls.

This model improves performance for the buyer. It does not change the structural relationship between buyer and seller, and it does not produce a shared record of the negotiation. The seller still receives a signal that a deal was struck at a given price; they do not have access to the negotiation process that produced it.

The distinction matters most when something goes wrong. When a discrepancy appears between what the buyer's system recorded and what the seller's system recorded, the resolution process currently relies on one party or the other providing evidence from their own internal logs. Neither party has an independent, shared record of the negotiation itself.

What does agent-to-agent deal negotiation actually require?

When agents represent both sides of a transaction, the structural requirements for the infrastructure supporting that transaction follow directly from the nature of agentic systems.

A shared deal record that both parties own. A deal negotiated between a buy-side agent and a sell-side agent cannot credibly live only inside one party's system. Both need a single authoritative record of what was agreed, when, and on what terms. A report sent from one platform to the other is not a shared record; it is one party's account of events.

Bilateral privacy-preserving negotiation. Each agent represents interests its counterpart is not entitled to see: budget constraints, yield floors, audience segment valuations. The negotiation mechanism needs to keep those inputs private while producing an outcome that is auditable by both parties after the fact. This is not a capability that a standard marketplace interface or a DSP optimisation layer provides.

Human approval gates that both parties can rely on. "Human control, agent scale" is a phrase Apostra uses, and the principle is broadly shared among credible participants in this space. The material question is where in the workflow the approval gates sit and who can enforce them. If the only approval mechanism is internal to one party's system, the other party has no way to verify that the control was actually applied.

A provenance record that survives after the deal completes. When a dispute arises, the parties need to establish what was agreed, which agent agreed to it, and on what basis. An audit trail internal to one platform is not an independent record. For agents acting with meaningful autonomy, the provenance record is the only mechanism by which the humans they represent can verify what was done in their name.

Marketplace discoverability and DSP-layer AI optimisation address none of these four requirements. They are real improvements to the advertising stack, but they are improvements to a structure that assumes human negotiation, not a replacement for it.

What should buyers ask when evaluating an agentic advertising claim?

A straightforward test applies when assessing any platform's agentic advertising claim. Call it the Deal Record Test, because all three questions turn on who owns and can verify the record of the deal.

First: who owns the deal record, and can both parties read it independently? If the deal record lives inside one platform and the counterparty receives a report, the system is optimising within one party's control rather than negotiating between two. Both models are useful. They solve different problems, and only one of them is agentic in the structural sense.

Second: where are the human approval rights enforced, and which party controls that enforcement? "Human in the loop" can mean anything from a mandatory approval gate that neither party's agent can bypass to an optional override that one platform controls unilaterally. The enforcing party and the mechanism are the details that matter.

Third: what happens when the agents disagree? In a genuine agent-to-agent system, disagreement is the normal case. A buy-side agent and a sell-side agent have different objectives. The infrastructure needs a resolution mechanism that produces an outcome both agents can act on, without deferring the disagreement to human teams to resolve manually.

Platforms that cannot answer all three questions concretely are offering AI-assisted advertising. That is valuable. It is not agentic advertising in the structural sense that the word implies when applied to negotiation between agents.

Where is this market going?

The Apostra rebrand is a useful signal about where the advertising technology industry believes the next category is being built. A company founded on carbon measurement and supply path transparency repositioning entirely around agent-ready advertising indicates that the discoverability and marketplace layer is now a recognised destination. Apostra's current product addresses that layer. Whether the company builds toward the deal negotiation infrastructure layer is a question the rebrand raises without yet answering.

The larger demand-side platforms are moving in a different direction: AI as an amplifier of existing buyer capabilities, operating within structures those platforms control. That is a defensible model for the segment of advertising where one party's agent negotiates against an automated price signal, rather than against another party's agent.

The category that remains underpopulated is the trust infrastructure layer: the shared deal record, the bilateral ownership mechanism, the provenance audit. This is the layer that agent-to-agent advertising requires and that neither marketplace discoverability nor DSP optimisation provides.

Alkimi's approach to deal negotiation infrastructure

Alkimi operates at the deal negotiation layer described above. Its DealSheet mechanism records the agreed deal in a shared record held outside either party's system, so the record both agents hold is the same record and neither can alter it unilaterally after agreement. Because that shared record, not one party's internal log, is the source of truth, a later dispute is settled against a version neither side controls. Negotiation inputs remain private to each party during the process and only the agreed terms are written to the shared record. Human approval is enforced as a gate on that record before it is committed, not reconstructed from reporting after the fact.

Alkimi does not replace the planning, inventory management, activation or measurement systems that buyers and sellers already operate. It handles the negotiation and governance layer between them: the part of the workflow that the existing advertising stack was not designed for, and that agent-to-agent advertising structurally requires.

Alkimi earns from transactions completing, not from either side winning a given deal, so its commercial incentive is a functioning, auditable negotiation rather than a favourable one for the buyer or the seller. A shared deal record held by a platform that is primarily a buy-side or sell-side tool cannot carry the same claim, because that platform profits from one side's outcome. The symmetry of the incentive, and the fact that the record is held by a party neither side controls, is what makes the bilateral ownership claim checkable rather than asserted.

Disclosure: this piece was produced by Alkimi's marketing function. Competitor positioning is sourced from publicly available company websites as of September 2026.