Most predictions about programmatic advertising in 2027 mix one thing the evidence supports with several it does not. The evidence supports a narrow claim: open frameworks for AI agents to trade advertising now exist, and negotiated deals are becoming something software can do, not only people. It does not support the louder claims that the open auction is ending, that humans are leaving the buying process, or that supply chain transparency has been fixed. The useful question for a buyer is not which forecast to believe but which indicator would show each one coming true.

TL;DR. Testing common programmatic predictions against evidence gives three verdicts: supported, partly supported and not supported. The claim that agent trading infrastructure is arriving is supported, because IAB Tech Lab announced AAMP in early 2026 and the first agent-to-agent buy using AdCP ran in October 2025. The claims that agents will replace the auction or remove human approval are not supported by how either framework is designed. Claims about transparency and protocol convergence are partly supported at best. Each prediction can be checked against an indicator a buyer can observe in their own data or in public registries.

How should a programmatic prediction be tested?

A programmatic prediction should be tested by separating it into three parts: the claim, the mechanism that would make it true, and an indicator a buyer could observe if it were happening. This piece calls that the claim, mechanism, indicator test. A prediction with no plausible mechanism is a slogan. A prediction with a mechanism but no observable indicator cannot be checked, so it should not drive budget decisions.

The broader picture of where programmatic is heading is covered in What is the future of programmatic advertising?. This piece takes the narrower job of checking individual forecasts that circulate in trade coverage and vendor decks.

Will AI agents replace the open auction?

Not supported. The open auction prices individual impressions in milliseconds across very large volumes of inventory, and no negotiation process, human or software, does that job better. An AI buying agent works over a longer horizon: it compares offers, negotiates terms and manages a budget across many transactions. That makes an agent a different transaction type, not a faster auction.

The mechanism that would make the prediction true does not exist yet: something that negotiates each long-tail impression at auction speed and cost. The indicator to watch is the split of spend in a buyer's own DSP reporting between open auction and deal IDs. If agents were replacing the auction, the open auction share would fall across all inventory tiers, not only premium.

Is agent-to-agent buying infrastructure actually arriving?

Supported. IAB Tech Lab announced AAMP, its Agentic Advertising Management Protocols, in early 2026 as the umbrella for all of its agentic work. AAMP publishes reference buyer and seller agents, the OpenProposal specification for how publishers describe products and packages in RFP responses, and a registry example for agent discovery. Separately, AgenticAdvertising.org reports that the first agent-to-agent media buy using AdCP, the Ad Context Protocol, ran on 16 October 2025.

Infrastructure is not adoption. The indicators that separate the two are concrete: how many of a buyer's existing sellers list an agent in the IAB Tech Lab Agent Registry, and how many return RFP responses in a standard format such as OpenProposal rather than a bespoke spreadsheet. Both can be counted today.

Will humans be taken out of media buying?

Not supported by the design of the frameworks. The AdCP documentation describes a governance layer in which a buy that exceeds an agent's authority, such as a budget above an auto-approval threshold, is escalated to a person who can add conditions before approving. AgenticAdvertising.org calls the principle "Embedded Human Judgment".

What is changing is where the human sits. For example, in a hypothetical £40,000 quarterly campaign, a trader who once approved every insertion order might instead approve the mandate at the start, sign off any single commitment above £10,000, and review exceptions. The indicator to watch is whether buying contracts and internal workflows start naming approval thresholds explicitly. The guide on how to write a media buying mandate sets out what those thresholds cover.

Has programmatic supply chain transparency been solved?

Partly supported, and only at the premium end. The ISBA and PwC follow-up study published in January 2023 found an impression match rate of 58% and an unattributable "unknown delta" of 3%, a large improvement on the original study. ISBA cautioned that the study covered the premium end of the market and should not be read as representative of the wider ecosystem.

Transparency tools have also improved who can be identified, not what was agreed. ads.txt (2017) lets publishers declare authorised sellers, and sellers.json (2019) lets intermediaries disclose who they pay. Neither records the commercial terms of a deal. The indicator to watch is whether contracts give the buyer log-level data and a record of agreed terms that the seller holds identically.

Will AAMP and AdCP converge into one standard?

Partly supported as a possibility, unsupported as a forecast. AAMP comes from IAB Tech Lab and builds agent capability onto existing standards such as OpenRTB and the Deals API. AdCP is published by AgenticAdvertising.org and defines tasks across the campaign, running over MCP or A2A. The two overlap in purpose, and whether they converge, coexist or compete is still open.

The indicator is vendor behaviour, not announcements. Watch which framework a vendor names, which components or version it supports, and which deal types those cover. Vendors claiming support for both at specific versions would suggest coexistence. Silence on versions suggests marketing.

Will new standards spread quickly?

Not supported by history. Programmatic standards have spread over years rather than months. The Media Rating Council issued its viewable impression guidelines in 2014, and viewability became a routine contract term only gradually after that. ads.txt took years to reach broad adoption after its introduction in 2017.

Agent standards may move faster because the parties building agents are also writing the frameworks. The indicator is whether agent terms start appearing in standard buying contracts rather than in pilots and press releases.

What should buyers do with these predictions?

Buyers should replace forecasts with a short watchlist they can check each quarter: open auction versus deal ID share, counterparties listed in the Agent Registry, the format of incoming RFP responses, approval thresholds in contracts, and access to a record of agreed terms. None of these requires a view on what 2027 will look like. Each shows what is happening in the buyer's own market.

The last item is where Alkimi, an agentic advertising marketplace working within IAB Tech Lab's AAMP framework, concentrates: buyer and seller agents agree terms on a DealSheet, a shared deal record both parties hold, while humans keep approval rights at the deal level.

Frequently asked questions

What is the future of programmatic advertising?

The evidence points to auctions continuing to clear most open-web impressions while a growing set of deals is negotiated by software agents under human-set mandates. How far that shift goes depends on adoption indicators that buyers can observe, not on forecasts.

Will AI agents replace DSPs?

There is no evidence for it. DSPs price auction impressions at speed. Agents negotiate and manage deals over longer horizons, so the two do different jobs.

Which programmatic predictions are best supported?

The arrival of agent trading infrastructure. IAB Tech Lab announced AAMP in early 2026, and AgenticAdvertising.org reports the first agent-to-agent buy using AdCP in October 2025.

How can a buyer test a programmatic forecast?

Use the claim, mechanism, indicator test: state the claim, identify what would make it true, and name a metric the buyer can observe. A forecast without an observable indicator should not drive spend.

Is there one agreed standard for agentic advertising?

No. AAMP from IAB Tech Lab and AdCP from AgenticAdvertising.org are developing in parallel, and their long-term relationship is still open.

Programmatic Advertising in 2027: Which Predictions Does the Evidence Support?