TL;DR: AI is shifting programmatic advertising from rule-based automation to reasoning-based decision-making, with consequences for how inventory is bought, how deals are structured, and how campaign performance is governed.

From Automation to Reasoning

Programmatic advertising was always automated, but early automation was mechanical. Rules were set by humans: bid this amount for this audience segment, cap frequency at this level, exclude these sites. The system executed the rules.

AI changes the nature of automation. A reasoning-based system does not simply follow rules. It evaluates conditions, weighs options, and selects actions based on objectives. The difference is material. A rule-based system performs consistently within its rules. A reasoning-based system can identify when the rules are producing poor results and adapt.

What AI Changes in Programmatic Buying

Targeting

AI-driven targeting models analyse behavioural and contextual signals at a granularity that rule-based segmentation cannot match. Rather than assigning users to broad demographic or intent categories, AI identifies patterns across large datasets and finds the specific combinations of signals that predict a desired outcome.

This has practical consequences. Campaigns can reach smaller, higher-value audiences rather than large, less relevant ones. Cost per outcome tends to fall even when cost per impression rises.

Bidding

AI bidding systems evaluate each auction opportunity against the current state of a campaign: remaining budget, pacing against delivery targets, historical performance of similar placements, and the predicted value of the impression. Human-set rules cannot account for this combination of variables in real time.

The result is more efficient spend allocation. Budgets concentrate on the impressions most likely to deliver against campaign objectives rather than spreading evenly across all available inventory.

Fraud and Brand Safety

AI is increasingly used to detect patterns associated with invalid traffic and brand safety risks before an impression is served, not after. This shifts the detection model from reactive review to proactive filtering.

Reporting and Attribution

AI models can process attribution data across large, complex customer journeys and identify which touchpoints genuinely influenced outcomes. Multi-touch attribution models built on AI are more accurate than last-click or rule-based models, though they require clean data inputs and careful validation.

The Shift to Agentic Advertising

AI in programmatic has so far been applied to execution: bidding, targeting, and delivery optimisation. The structural shift underway is towards agentic advertising, where AI moves upstream into deal-making.

In agentic advertising, a buy-side AI agent negotiates directly with a sell-side AI agent to agree the terms of a deal before any impression is served. This is not bidding in an auction. It is a negotiation, with offers, counter-offers, and a documented agreement.

The Agent Communication Protocol (AdCP), published by AgenticAdvertising.org, defines the message formats for these agent-to-agent negotiations. The AAMP, published by IAB Tech Lab, provides broader governance standards for how agentic transactions are structured and documented.

Alkimi's role is to provide neutral infrastructure for these negotiations. The DealSheet is the mechanism by which both agents hold a shared, private record of an agreed deal. The marketplace maintains the audit log and the approvals workflow, while the rest of each party's stack remains unchanged.

Governance and Accountability

One of the most significant challenges in AI-driven programmatic advertising is governance. When an AI system makes a decision that produces a poor outcome, the audit log needs to show what information the system had, what decision it made, and why.

This is not a theoretical concern. Regulatory pressure on automated decision-making is increasing, and clients are asking harder questions about where their money is spent and what decisions were made on their behalf.

Agentic advertising frameworks address this by design. Every negotiation between agents produces a documented record. Every deal has an audit log. Every action taken by an agent is bounded by parameters set by a human principal. This is the governance model that separates responsible agentic advertising from uncontrolled automation.

Summary

AI is not changing programmatic advertising at the margins. It is changing the nature of what programmatic means: from mechanical automation to reasoning-based decision-making, and from execution-only systems to agents that negotiate the terms of trade.

US advertising teams that understand this shift will be better positioned to evaluate the tools, vendors, and protocols that will define the next phase of the industry.