Artificial intelligence is changing programmatic advertising at every stage of a buy: it helps plan campaigns, prices bids in milliseconds, produces and checks creative, models what media actually caused and, newest of all, lets software agents negotiate deal terms with other agents. The older changes make existing processes faster and more granular. The newest one changes the transaction itself, so the central question shifts from what the software can do to who answers for what it agreed.

TL;DR. Machine learning has priced and paced programmatic bids for years, so AI in programmatic is not new. What changed in 2026 is that agents moved upstream into planning and deal-making, supported by open standards such as IAB Tech Lab's AAMP and the Ad Context Protocol (AdCP). Creative and measurement are being reshaped too, but both still depend on human approval and clean data, and the media buyer's job is moving from executing transactions to governing the mandates agents work within.

One concrete example of the newest stage: on Alkimi, buyer and seller agents are benchmarked in Concourse, built and tested in the Gym against counterparty agents before any real money moves, and then negotiate on the Marketplace, where a person approves each deal and the agreed terms are recorded on a DealSheet that both sides hold.

How is AI different from the automation programmatic already had?

Programmatic has always been automated, but most of that automation followed rules a trader wrote: bid this much for this segment, cap frequency here, exclude these sites. A rule-based system performs consistently inside its rules and cannot tell when the rules are producing poor results. A model-based system weighs many signals against an objective and adapts.

AI itself is not new to the ad stack. "AI in advertising is typically framed as a recent development, but that's just wrong," Anthony Katsur, CEO of IAB Tech Lab, said in a keynote that IAB Tech Lab later published as a history of AI in advertising, describing how a system "evaluates a user, assesses context, generates a price, and executes a transaction" in roughly 400 milliseconds.

The structural change is newer. Agents now work from a mandate, a defined scope of authority set by a person, hand decisions outside that scope back to a human and keep a record of what they decided. That move, from a person deciding with software assistance to software deciding within limits a person sets, sits behind every stage described below.

How is AI changing programmatic planning?

Planning has been the most manual stage of programmatic, and it is where agents are newest. IAB Tech Lab observes that a media team may have time to evaluate proposals from 10 or 20 publishers for a single campaign, while hundreds or even thousands of others never see the brief. Agents can read and compare far more proposals than a planner can, provided sellers describe their inventory in a format an agent understands.

That is the job of OpenProposal, the specification at the centre of AAMP 3.0, which IAB Tech Lab announced on 22 September 2026. Sellers publish packages, buyer agents discover and compare them against a declared brief, and agents can negotiate custom terms while keeping a reference to the original offer. "What's new in AAMP 3.0 is that the reasoning travels with the proposal," Mike Treon, head of CTV strategy at PMG, said in the announcement.

Planning agents are also reaching television. In the agentic initiative between NBCUniversal and the agency RPA, the buy-side agent starts the RFP process by sharing campaign goals and requesting proposals for linear and streaming inventory, AdExchanger reported. RPA's Lisa Herdman told the outlet that agentic tools are starting to help planners identify new audiences.

How is AI changing programmatic bidding and pricing?

Bidding is the most mature use of AI in programmatic. After the industry moved to first-price auctions, which Google Ad Manager began rolling out to all its partners in September 2019, bid shading models became standard for estimating the lowest price likely to win an impression. Modern bidding models evaluate each opportunity against remaining budget, pacing, past performance of similar placements and the predicted value of the impression, a combination no hand-written rule set can weigh within an auction's time limit.

Pacing has moved from fixed daily caps to continuous adjustment across dayparts, inventory types and channels. Targeting has moved from broad demographic segments towards models that find the combinations of behavioural and contextual signals that predict an outcome, which matters more as third-party identifiers become less available. Budget allocation increasingly follows the forecast likelihood of an outcome rather than yesterday's results alone.

Brand safety and fraud detection have shifted from reactive to proactive. Contextual classification models assess the content of a specific page rather than applying keyword blocklists. Invalid traffic detection increasingly filters suspicious requests before a bid rather than in a post-campaign report.

Agents are now automating the setup around the bid as well. NBCUniversal's chief product officer, Ryan McConville, told AdExchanger that a DSP can automate buying a curated deal ID, but setting up the deal ID, choosing what inventory to include and handling the logistics has been manual work that agentic tools are good at automating.

How is AI changing programmatic creative?

Generative AI is now a routine part of ad production. Research by IAB and Sonata Insights, published in January 2026, found that 83% of ad executives say their company has deployed AI in the creative process, up from 60% in the 2024 study. Dynamic creative optimisation, which assembles variants from components and shifts delivery towards better performers, can now draw on components the models generate themselves.

The constraint has moved from production to approval. The same IAB research found that 82% of ad executives believe Gen Z and Millennial consumers feel positive about AI-generated ads, against 45% of those consumers who say they do. IAB's AI Transparency and Disclosure Framework takes a risk-based approach, requiring disclosure only when AI materially affects authenticity, identity or representation in ways that could mislead consumers.

Agents also speed up creative operations. FreeWheel's chief product officer, David Dworin, told AdExchanger that agents can approve creative in seconds, which matters in live sport, where an ad whose creative has not been approved in time typically does not run.

How is AI changing programmatic measurement?

Measurement is changing in two directions at once. Modelled measurement is becoming more accessible: Google's Meridian is an open-source marketing mix model, and its GeoX integration lets advertisers calibrate the model with the results of geo experiments. Models of this kind are more flexible than last-click attribution, but they are only as reliable as their inputs and need validating against controlled tests.

Measurement terms are also moving into the deal itself. The AAMP 3.0 lifecycle ends with a report stage, so that results can be "evaluated against the original objectives and commitments of the plan". AdCP's accountability documentation defines performance standards, the vendor that counts the billing metric and make-good remedies as structured, machine-readable fields that agents negotiate before budget is committed.

A completed agent buy is not necessarily a well-protected one. In Concourse Bench v1, 55 of 96 registered attempts by AI buying agents ended in a completed buy, yet 95 of the 110 contracts signed in those buys promised a viewability floor below 70%. Incrementality is a separate question again, because a promised floor says nothing about what a campaign caused, so holdout tests still matter.

How are AI agents changing programmatic deal-making?

Negotiation is the newest stage AI has reached. In an auction, a price is won; in an agent-negotiated deal, a buy-side agent proposes terms, a sell-side agent accepts, counters or declines, and the terms are agreed. That is a different transaction type, not a faster auction.

The first agent-to-agent media buy ran on 16 October 2025, according to AgenticAdvertising.org, which publishes AdCP, with real money, inventory from LG Ads and AI agents handling creative, targeting and approval with human oversight. On the IAB Tech Lab side, AAMP 2.0 extended the agent framework through the Deals API to Programmatic Guaranteed, Preferred Deals, private marketplaces, the open programmatic marketplace and linear GRPs. PubMatic launched AgenticOS on 5 January 2026 so that agents can plan, transact and optimise programmatic campaigns, interpreting intent expressed through protocols including AdCP.

Agents add a layer rather than removing the one underneath. McConville told AdExchanger that agents sit on top of programmatic rather than replacing it, and in the NBCUniversal and RPA workflow the outlet describes, the campaign executes automatically only "once humans on both sides of the deal sign off". IAB Tech Lab says OpenProposal works without requiring DSPs, SSPs or other technologies to change their existing use cases; how the two models will coexist is covered in a separate analysis of the future of programmatic advertising.

What has changed in programmatic advertising in 2026?

In 2026, agentic buying moved from demonstrations into live campaigns, pilots and formal standards. A separate list of ten sourced AI trends in programmatic covers curation, CTV and agent identity in more depth.

16 October 2025: the first agent-to-agent media buy runs, according to AgenticAdvertising.org.
Early 2026: IAB Tech Lab announces AAMP as the umbrella for its agentic advertising work.
5 January 2026: PubMatic launches AgenticOS for agent-to-agent advertising.
6 January 2026: NBCUniversal, RPA, FreeWheel and Newton Research announce agentic buying across linear and digital, with a first execution covering a brand's Q1 2026 investment, including live football playoff games.
January 2026: IAB publishes its AI ad gap research and its first AI Transparency and Disclosure Framework.
4 March 2026: IAB Tech Lab introduces its Agent Registry, which associates each registering company with its GPP ID for validation.
21 April 2026: IAB Tech Lab argues that agentic direct-sold campaign state must be "machine-readable, deterministic, and canonical" and proposes an AgenticDirect extension to OpenDirect.
22 September 2026: IAB Tech Lab announces AAMP 3.0 with OpenProposal, bringing agents into the RFP and proposal stage.
October to November 2026: OpenProposal public comment, originally set to close on 22 October 2026, is extended to 10 November 2026.

What governance gap does AI create in programmatic?

Agents can commit to deals faster than existing reconciliation processes can check them. Manual reconciliation was designed for the deal volumes human traders negotiate, and it cannot keep up when discrepancies emerge across many agent-made deals.

IAB Tech Lab argues that, for agentic direct-sold advertising, campaign state belongs in a structured data model "that both buyer and seller agents read from the same source of record". The practical response is machine-readable deal governance: agreed terms both sides can query, discrepancy flags raised automatically and human attention reserved for genuine exceptions.

Identity and accountability belong to the same gap. IAB Tech Lab's Agent Registry lets companies describe each agent's purpose and maturity, and AdCP's Embedded Human Judgment principles state that "responsibility cannot be delegated to software". One staged approach is earned autonomy, in which an agent moves through five levels as it proves reliable: Observe, Recommend, Draft, Human-approved action and Bounded automatic action.

What does AI not change in programmatic advertising?

AI does not change the scarcity of premium inventory, whoever or whatever is buying it. It does not remove the need for brand safety governance, because agents apply the rules but people set them and answer for the outcome. And it does not break the principal-agent chain: the organisation that deployed a buying agent remains accountable for what that agent commits to.

AI does not make programmatic transparent by default either. Model-driven decisions can be harder to trace than written rules. Transparency improves only when agreed terms are recorded in a form both buyer and seller can check against delivery.

How is the media buyer's role changing?

Agents take the high-frequency, rule-bound tasks: bid adjustments, pacing, brand safety filtering and budget allocation across inventory types. Oversight changes shape, because a person reviews the decisions an agent escalates rather than every decision it makes. The work that stays human is what a mandate cannot yet specify: interpreting a strategic brief, evaluating new partners, creative direction and deciding what the mandate itself should say.

Testing becomes part of the job. Before an agent handles a real budget, buyers need to check it against the outcomes that matter to them, such as contract protections, not completion alone, as set out in a separate guide on how to test an AI media buyer before deployment.

What does this mean for teams buying and selling programmatic media?

The common thread across every stage is accountability: which agent acted, who approved the commitment and which record both sides trust.

Alkimi, an enterprise platform for agentic advertising, is built around those questions. Its Marketplace is neutral, works within IAB Tech Lab's AAMP framework today, is being built to be AdCP compatible, uses AAMP identity checks and keeps human approval rights at the deal level. Each agreement is recorded on a DealSheet, written at agreement, timestamped, held by both parties and retrievable by either side independently, and the company took part in the OpenProposal v3.0 public comment period, filing comments and change requests.

Frequently asked questions

How is artificial intelligence changing programmatic advertising?

AI now helps plan campaigns, price bids, produce and approve creative, model measurement and, newest of all, negotiate deal terms through software agents. People still set the objectives, the limits and the approvals that commit money.

Is AI bidding new in programmatic advertising?

No. Machine learning has priced and paced programmatic bids for years, and bid shading became standard after the move to first-price auctions.

Will AI replace media buyers?

AI takes over high-frequency decisions such as bid adjustments and pacing. Setting the mandate, interpreting the brief and approving consequential commitments stay with people.

What is the difference between AI optimisation and agentic advertising?

AI optimisation improves decisions inside an existing auction, such as how much to bid. Agentic advertising has software agents act for a buyer or seller, including agreeing deal terms with another party's agent, within limits a person has set.

Which standards cover AI agents in programmatic advertising?

The two main open frameworks are AAMP, IAB Tech Lab's Agentic Advertising Management Protocols, which added OpenProposal in version 3.0, and AdCP, the Ad Context Protocol published by AgenticAdvertising.org. They come from different bodies and are developing in parallel.

How Is Artificial Intelligence Changing Programmatic Advertising?