Agentic advertising is frequently described as the next evolution of programmatic. The framing is understandable: both involve technology mediating the purchase of advertising inventory, and both operate at a scale and speed that human buyers could not achieve manually. But the framing obscures a structural distinction that matters for how you build, govern, and evaluate these systems. Programmatic is a buying mechanism. Agentic buying is a buying model. They are not the same thing, and understanding the difference matters more than most industry commentary suggests.

What is programmatic advertising?

Programmatic advertising is the automated, data-driven purchase of advertising inventory through technology. The word "programmatic" refers to the use of software to execute transactions that were previously handled through direct human negotiation and insertion orders. A buyer sets parameters, the technology executes against those parameters, and inventory is purchased, served, and reported on without requiring a phone call or a manually drafted contract for each placement.

The technology infrastructure that makes programmatic work includes demand-side platforms (DSPs), supply-side platforms (SSPs), data management systems, ad servers, and verification services. These systems communicate in real time to evaluate and transact on individual ad impressions at speeds measured in milliseconds. The human buyer's role in this process is to set the strategy: define the audience, set bid parameters, determine budget allocation, specify quality requirements, and review the results.

The key point about programmatic is what it automates and what it does not. It automates the execution of a pre-specified strategy. The human still decides what to buy, for whom, and why. The machine executes the transaction at a speed and scale that the human cannot match manually. Decision authority remains with the human. Execution authority has been delegated to the technology.

What changes when agentic buying enters the picture?

Agentic buying changes what the machine does. In a programmatic model, the machine executes a strategy the human has defined. In an agentic model, the machine reads a brief, determines a strategy, negotiates the terms of individual deals, and assembles a portfolio. The decision layer moves from the human to the model.

This is not a natural evolution of programmatic. It is a structural change in who, or what, makes the consequential decisions. The underlying infrastructure, the pipes through which inventory is transacted, can in principle be the same. But the governance layer and the accountability model are categorically different. When a human sets programmatic parameters and the results are poor, the human reviews and adjusts the parameters. When an AI model assembles a portfolio and the results are poor, the question of what went wrong, and who is accountable, requires a different kind of analysis.

The infrastructure can be the same. The governance cannot be. That is the distinction most vendors describing agentic buying as "programmatic 2.0" are eliding.

What did programmatic change, and what did it leave unchanged?

Programmatic changed how transactions are executed. It did not change who makes buying decisions. A media planner using a DSP in 2024 is still responsible for the strategy: which audiences to target, which channels to prioritise, what budget to allocate. The DSP executes that strategy more efficiently than manual insertion orders allowed. The accountability model is unchanged from pre-programmatic direct buying: the planner is responsible for the plan, and the technology is accountable for execution fidelity.

Programmatic did introduce complexity into the supply chain that pre-programmatic direct buying did not have. The intermediary layers that accumulated between buyer and publisher created opacity, raised questions about brand safety and fraud, and made it difficult for buyers to verify that their money was reaching the media they intended to buy. These are real problems, and they have produced real governance failures. But they are problems of execution, not of decision authority.

Agentic buying introduces a different kind of complexity. It shifts decision authority from a human to a model. That requires a different governance response: not just verification that the technology executed correctly, but verification that the decisions the model made were the right ones, within the right constraints, with the right approvals.

Why does the accountability model differ between programmatic and agentic buying?

In programmatic buying, accountability is relatively straightforward. The planner set the parameters. If the campaign did not deliver, the parameters were wrong or the execution was flawed. Both can be investigated from the available data: the bid log, the impression log, the delivery report. The human made the strategy decisions and is accountable for them.

In agentic buying, the model made the strategy decisions. Accountability requires that those decisions be legible after the fact. What did the model understand the brief to require? Which sellers did it approach, and why? What trade-offs did it make when assembling the portfolio? Where did it deviate from the brief's requirements, and what drove that deviation?

None of these questions can be answered from a bid log or a delivery report. They require a deal record that captures the model's decision-making at the transaction level. Without that record, accountability is nominal. The client can see what was bought and what it cost, but not whether the buying was well-governed or whether the model behaved within its brief.

For buyers coming to agentic from adjacent fields, what is the essential distinction?

Programmatic advertising is a transaction infrastructure. It answers the question of how to execute media purchases at speed and scale. It does not answer the question of what to buy or why. That remains a human judgement.

Agentic buying is a decision infrastructure. It provides the capability for an AI model to make buying decisions, within a defined brief and with defined human approval rights, rather than executing a strategy that a human has already decided. The underlying transaction pipes may overlap with programmatic infrastructure. The question of who decides is fundamentally different.

For buyers evaluating whether and how to adopt agentic technology, the relevant questions are not primarily about programmatic expertise. They are about brief quality (what parameters will govern the model's decisions), governance design (where will humans retain approval rights), and accountability infrastructure (what record will show that the model acted within its brief). These are governance questions, not technology questions, and the answers determine whether agentic buying creates value or creates risk.

What does Alkimi build, and where does it sit in this picture?

Alkimi builds the infrastructure for agent-to-agent media trading. That means the negotiation layer between buy-side and sell-side agents, the deal record that captures what was committed and on what terms, and the approvals workflow that ensures human oversight at defined decision points. What Alkimi does not replace is the transaction infrastructure underneath: the ad serving, the measurement, the planning tools, the inventory management systems that sit on both sides of a media transaction.

Programmatic is a buying mechanism. Agentic is a buying model. Alkimi builds the infrastructure for the latter, on top of the infrastructure the industry has already built for the former. The two are complements, not competitors. But treating them as equivalent, or describing agentic as simply "more programmatic," is a category error that will produce bad procurement decisions and worse governance.