The way advertising gets bought and sold has changed. AI agents are taking on tasks that have always required human coordination at every step. The workflows are shifting, and for them to work smoothly at scale, the ecosystem needs a common set of rules. The Ad Context Protocol is becoming one. In this article, we'll take a look at how it works, who is building on it, and what it means for the market.
There is a moment in every market shift when the infrastructure catches up to the idea. For programmatic advertising, that moment was OpenRTB. Before a shared protocol existed for real-time bidding, every exchange, DSP, and publisher was solving the same coordination problem in isolation. OpenRTB did not invent programmatic advertising. It made it scalable. It gave the industry a common language, and that common language unlocked a decade of compounding investment and growth.
Agentic advertising is at a similar inflection point. AI agents are beginning to take on tasks that humans have always handled in the advertising workflow: researching inventory, negotiating terms, activating campaigns, and reporting on delivery.
We believe agentic media buying represents a fundamental shift in how advertising will be transacted, and the industry is responding by developing protocols that allow these agents to communicate in structured, standardized ways across different platforms and systems. While alternatives exist, we expect the ecosystem to support more than one standard as the market matures. Sigma Software is actively engaged across this space, and for this article, we are focusing on AdCP — the Ad Context Protocol. It has gathered an active community around, and we are already seeing significant spend flow through its agentic pipelines.
What AdCP actually solves
The promise of AI in advertising has mostly been realized at the optimization layer: smarter bidding, better targeting, faster reporting. What has not yet been addressed at scale is the deal layer, the human-intensive process of how advertising relationships are initiated, structured, and activated in the first place.
That process has four stages, and each is slow by design:
- Discovery
A buyer identifies a publisher, evaluates inventory quality, and determines whether there is a fit. Today, this means relationship networks, sales calls, and significant manual research on both sides - Negotiation
Terms, pricing, audience guarantees, and creative specifications get worked out across teams and timelines. A process that should take hours routinely takes weeks. - Activation
Once terms are agreed, the campaign still has to be stood up in both buying and selling systems. This step requires technical coordination that does not always go smoothly and often adds more time. - Reporting
Validating that what was agreed is being delivered remains largely manual for direct deals, and reconciliation across systems is a persistent operational burden.
AdCP is designed to compress these four stages. It enables AI agents on both buy and sell sides to handle them directly via structured machine-to-machine communication, without a human needing to facilitate each step. A Buyer Agent can discover available inventory, evaluate it against campaign objectives, negotiate terms within pre-authorized parameters, and trigger activation in seconds rather than weeks.
None of this means removing humans from the process entirely, at least not yet. Agentic workflows are designed with human-in-the-loop controls built in, and most organizations will want to define specific points where human approval is required before the system proceeds. A creative execution might need sign-off before it goes live. Custom pricing or audience packaging might require a human to review terms before they are accepted. If a campaign is underdelivering, an Ad Ops team member can step in to make a call directly or review a recommendation the agent has already prepared. The practical reality is that agentic advertising shifts where human judgment is applied, not whether it is applied. The repetitive coordination work gets handled automatically. The decisions that carry real business or brand risk stay with the people accountable for them.
This is not a speculative capability. It is being built and tested now.
Money is real
The most important signal in the agentic advertising market right now is not the protocol. It is the direction of the money.
Real spend is beginning to flow through agentic pipelines. Campaign budgets are moving into agentic workflows, and while the scale is still early, the trajectory is clear.
In a recent Prebid webinar, Sigma Software ran an industry survey to map where companies stood on testing agentic advertising, how much spend had already been executed, and what the mix of participants looked like across the ecosystem. The findings confirmed what many in space are starting to feel: this is no longer a conversation about whether agentic advertising works. If you want to dive deeper into the survey results, the full Prebid webinar is available to watch.
The ecosystem is evolving quickly, and the organizations building now are not running experiments. They are making infrastructure decisions.
And the range of participants is expanding. DSPs and SSPs are integrating agentic capabilities into their platforms. Agencies are standing up Buyer Agent infrastructure to automate how they plan, negotiate, and activate media. Data companies are building tooling to make their signals accessible to agentic buyers. Independent vendors across the ecosystem are evaluating where they fit and how to remain relevant as the workflow changes around them.
For publishers, the shift is meaningful but gradual. Direct deals are still largely relationship-driven today, who you know, who your sales team can reach, and which buyers are willing to take a meeting. That is not disappearing overnight. But agentic buying is already making those relationships easier to initiate and faster to execute, and over time, discovery itself will become increasingly automated.
As agentic discovery matures, a Buyer Agent evaluating inventory may not need a prior relationship to find a publisher at all. That is a real opportunity for publishers who have historically been locked out of direct demand, including the long tail of publishers who rarely see large direct budgets simply because they lack the sales infrastructure to compete for them.
AdCP is more than Buyer and Sales agents
When people talk about agentic advertising, the conversation tends to center on the transaction: a Buyer Agent talking to a Sales Agent to execute a deal. That is an important part of the picture, but the agents doing the visible transactional work depend on a layer of supporting agents running underneath them. Those supporting agents are what make agentic advertising function reliably at scale:
- Signal Agents are one of the more critical pieces. These are agents built by data companies and data platforms to surface audience intelligence, contextual signals, and identity data in formats that agentic buyers can actually consume and act on. In a traditional workflow, a buyer pulls data from a DMP or cleans room manually and applies it to a plan. In an agentic workflow, Signal Agents make that intelligence available in real time, at the point of decision. The quality and breadth of signal available to a Buyer Agent directly affect the quality of the decisions.
- Governance Agents address a different but equally important need. Brand safety, verification, and compliance requirements do not go away in an agentic environment. If anything, the speed of agentic transactions makes them more important, because there is less time for manual review before an ad goes live. Governance Agents, built by verification vendors and brand safety companies, can operate continuously across agentic transactions, enforcing standards and flagging issues in real time rather than surfacing them in a post-campaign report.
- Creative Agents are earlier in development, but the underlying need is clear. Agentic buying can move faster than traditional creative production allows, and closing that gap requires agents that can adapt, personalize, or assemble creative executions against the parameters being negotiated in real time.
The Buyer Agent and Sales Agent handle the commercial relationship. The Signal, Governance, and Creative Agents make that relationship worth having.
Open Protocol is still being written
AdCP is not a finished specification. It is an active and evolving standard, and that is worth understanding clearly.
The protocol currently addresses core transactional workflows around discovery, negotiation, and activation. That is meaningful progress. But significant areas of the advertising workflow remain largely outside the current specification.
Planning, the upstream process of how campaigns are scoped and budgeted before deal execution begins, is one. Publisher onboarding, the process by which inventory becomes discoverable and credentialed within agentic systems, is another. Cross-channel reporting standards for agentic transactions are still being developed.
Now is the opportunity to shape the future of agentic media buying. The organizations contributing today are the ones shaping what the standard becomes, and that is a meaningful position to hold.
For product leaders and senior executives across the ecosystem, the more pressing question is not whether agentic advertising will affect your business. It already is. The question is how your organization engages with it and how quickly you move.
The evaluation process looks different depending on where you sit.
- Publishers need to ask whether their inventory is legible to agentic buyers.
- Agencies and trading desks need to assess what Buyer Agent capabilities they build internally versus adopt from the market.
- Data and verification companies need to determine how their existing products translate into agentic formats and which agent types represent their best point of entry.
- DSPs and SSPs are making infrastructure decisions right now about how deeply to integrate agentic workflows into their core platforms.
Across all these cases, the advice is the same: the people and organizations you work with need to understand the standards as they stand today, have visibility into where they are heading, and be capable of supporting the specific use cases relevant to your business. The agentic advertising ecosystem is evolving quickly enough that working with someone who is a step behind is a meaningful risk.
The programmatic market did not wait for OpenRTB to be complete before moving. It developed with the protocol, and companies that engaged early gained structural advantages that took years to replicate.
Agentic advertising follows a similar path. Thus, the organizations that make decisions now about what to build, who to partner with, and how to position themselves will be ahead of those who revisit the question in 18 months.
Sigma Software is a founding member of the Agentic Advertising Organization and an active contributor to the AdCP working groups. Olga Paramonova is Vice Chair of the Agentic Task Force at Prebid.
We work with publishers, buyers, and ecosystem vendors across the agentic advertising stack. Come meet us at Cannes Lions this June to discuss what we can do for your organization.
