Programmatic CPMs have been under pressure for years. More inventory, better floor prices, and additional SSP relationships haven’t made a dent, because these tactics skirt around the root of the issue.

Declining CPMs are, in part, a result of publishers gradually ceding control over how their inventory is packaged, priced, and sold. Sell-side curation is how publishers take that control back.

What Is Curation?

Curation is the practice of packaging inventory and audience data into deal products that buyers can target directly via a Deal ID within their preferred DSP. Sitting between publishers and advertisers, a curator assembles inventory based on audience signals, contextual criteria, or quality thresholds. Buyers target the Deal ID like any other inventory.

Historically, curation happened inside the DSP. Buyers and intermediaries controlled the logic behind curated packaging; the publisher was rarely, if ever, involved in decisions about how inventory was selected, combined, and priced. 

But curation can happen on the sell side before the auction begins. That shift gives publishers more control over how their inventory is packaged and ensures buyers have access to premium inventory.

Benefits of Sell Side Curation

When curation happens on the buy side, the packaging intelligence and the margin from that activity accrue to the intermediary. Publishers provide the inventory data but have limited visibility into how it's assembled, what signals are applied, or what the buyer ultimately paid.

Sell-side curation reverses that dynamic, which compounds into benefits on both sides of the ecosystem.

For Sellers

• Control over targeting logic: Publishers define the audience, apply quality thresholds, and set pricing themselves, rather than leaving those decisions to an advertiser or intermediary.
• Margin retention: The value generated by packaging and curating publisher audiences stays with the publisher.
• Real-time decisioning: Sell-side curation can happen in as little as 10 milliseconds versus 100 or more on the buy side. This processing speed allows smarter pricing before the auction and continuous reoptimization as attributes change.

For Buyers

• Cleaner supply paths: Fewer intermediaries between the buyer and the impression means better data signal integrity and less bid duplication.
• Greater transparency: Curated supply gives buyers clearer visibility into where impressions originate, how inventory is packaged, and which signals are used.
• More consistent outcomes: Pre-qualified, high-signal supply reduces the need for heavy DSP-side filtering and the cycles that waste budget with costly guesswork.

The end result: Publishers capture more of the value their premium audiences generate, and buyers get access to cleaner, more reliable inventory.

Hallmarks of a Well-Structured, Curated Deal

The deals that command premium CPMs go further, combining audience precision with quality signals that give buyers confidence in who they're reaching and where the impression runs. Defining characteristics include:

• A well-resolved, enriched audience: Layering in third-party demographic, behavioral, and intent signals adds audience depth and specificity, giving buyers the confidence to bid aggressively.
• Quality controls applied before audience mapping: Applying quality filters before matching users to a segment means every impression must meet brand and environment standards, resulting in better-fit audiences than traditional curated deals can offer.
• Flexible deal packaging: One audience can power multiple deal types, like broad reach and premium quality placements, without rebuilding segments.
• Multi-SSP activation under a single Deal ID: Activating the same audience package across SSPs and other destinations drastically reduces the manual overhead of building and managing deals individually per SSP.

Agentic Workflows Enable a Curation Revenue Strategy

Building one well-structured curated deal is manageable, if time-consuming. Accomplishing it across a growing library of deal products, and keeping each updated as audience data accumulates and campaign requirements shift, is where most teams run out of steam.

Sell-side agents help publishers mitigate that gap by expanding audience addressability. Agents can draw from enriched audience data and available inventory to propose and refresh curated packages in real time. By cross-referencing signals and surfacing combinations agentically, sell-side agents reduce the manual effort and time commitments required to create and sustain these curated audiences.

Agents can also make curated deal products automatically available to buyer agents searching for relevant inventory, based on parameters the publisher has already defined. 

With agents handling the operational weight of building and maintaining deal products, curation stops being a reactive exercise and starts functioning as a continuous revenue strategy. The publisher's audience intelligence compounds with every campaign rather than getting rebuilt from the ground up each time.

Curate Your Premium Inventory On Your Terms

Sell-side curation is a fundamental shift in where publisher value sits in the programmatic supply chain. Instead of having limited visibility into how their inventory is packaged and sold, publishers gain active control over more layers of the deal. The publishers who build agentic, audience curation infrastructure now are the ones who will compete for premium demand on their own terms, rather than waiting to see how buyers choose to package their supply.

Optable brings the full stack together: enriched audiences, quality controls, multi-destination activation, and the agentic layer that makes it sustainable at scale. 

Ready to build a scalable curation strategy? Schedule a demo today.

When a buyer requests a niche audience, publishers can usually intuit if they have users who match the profile. But whether they can verify, size, and package them with enough precision to act on the deal depends on their data infrastructure.

Publishers have the right audiences; they just don’t always have the tools to confirm who fits, enrich the signal, and package them with confidence.

AI is already clearing that roadblock. By drawing on behavioral signals, contextual patterns, and third-party attributes alongside first-party data, AI tools can help publishers verify and size audiences that previously fell below the threshold of what publishers could confidently sell.

Where First-Party Data Alone Can’t Reach

First-party data is the most valuable signal a publisher owns, but it has structural limits that become apparent the moment an advertiser asks for precision at scale.

Most publishers can only identify up to 30% of their audience with their first-party data, according to Optable's State of Audience Data Monetization report with Digiday. The majority of traffic remains anonymous, which means addressable segments are already working from a reduced starting point before any targeting requirements are applied. The qualifying pool shrinks further when an advertiser adds specificity, such as a particular income bracket, a purchase intent signal, or a geographic constraint.

The compounding effect on revenue:

• Niche segments that exist at small scale get passed on rather than proposed, creating revenue losses that never appear in deal reporting.
• Publishers who can't fulfill at scale water down audience requirements to hit volume targets, reducing the precision that made the segment worth buying in the first place.
• Campaigns that should be winnable go to competitors with larger or more enriched audience graphs.

How Small Audiences Can Punch Above their Weight Class

Even when an audience is small or your site traffic is down, you can monetize your audience effectively. The following approaches can expand what's addressable without compromising what made the original segment valuable.

Signal Enrichment

Enriching your identity graph with third-party signals is the most direct path to expanding what you know about your audience. First-party behavioral data tells you what someone did on your properties. Third-party enrichment tells you who they are, what they own, what they're shopping for, and where they are in a purchase cycle.

Optable's attribute taxonomy includes approximately 800 segments covering demographic, behavioral, transactional, and in-market intent signals sourced from partners including LiveRamp, Experian, and TransUnion. When layered onto a publisher's first-party graph, these signals transform a narrow behavioral audience into a richer, more addressable one.

A segment solely defined by content engagement can become a segment informed by content engagement as well as income range, household composition, purchase intent, and more. The enriched audience is a fundamentally more valuable product for advertisers and a larger qualifying pool for the publisher.

Lookalike Modeling

When a high-value audience segment is too small to fulfill a campaign, lookalike modeling identifies the shared behavioral patterns, demographic attributes, and intent signals to broaden the audience.

The approach works within a publisher's own data environment and through clean room collaboration with an advertiser or partner. In the latter case, the overlap between a publisher's first-party data and an advertiser's customer list becomes the seed audience. The model expands outward from there, finding similar users across the ecosystem without either party exposing raw data to the other.

In either scenario, the output is a modeled audience segment ready for activation that extends the reach of the original without sacrificing the precision that made it worth building.

AI-Enhanced Audience Discovery

Where enrichment adds depth and lookalike modeling adds reach, AI-enhanced audience discovery expands what the publisher can find within their own data.

Optable's Audience Agent can take a seed audience and propose expansion strategies based on the signals available in the enriched graph, relaxing income thresholds, broadening intent signals, substituting behavioral proxies for signals that are present in the seed but underrepresented at scale.

Each expansion path comes with a confidence ranking, giving the sales or planning team a clear view of which approaches maintain the original segment’s characteristics and which trade precision for volume. This would take a data analyst hours of manual permutation testing, but an agent can surface this information in minutes, with transparent reasoning at every step.

Expand Your Audiences Without Diluting Data Quality

A modeled or enriched segment that looks bigger but performs worse doesn't solve the addressability problem. It damages the advertiser relationship and hinders campaign success.

Confidence scoring on AI-proposed expansion paths gives teams the ability to evaluate which strategies preserve the intent and behavioral characteristics of the original seed before they propose anything to an advertiser. Trait indexing shows how an expanded audience compares to the original on every attribute in the graph, so publishers can demonstrate that a scaled-up segment still indexes strongly against the signals that made it worth buying.

Expanded Audiences Lead To Bigger Deals 

Expanding addressable reach affects which deals a publisher can compete for, not just how many they can fulfill. The impact shows up across four areas of the revenue operation:

• CPM and deal type: Larger, more precisely defined audiences qualify for higher CPMs and make publishers more competitive for programmatic guaranteed and direct-sold campaigns.
• RFP win rate: When a seed audience is too small to fulfill a brief, an Audience Agent proposes expansion paths in real time rather than passing on the opportunity.
• Budget access: Lookalike modeling and collaborative audience matching open budget categories that require audience precision to access.
• Compounding value: Every expansion that performs well builds the data story for the next campaign, strengthening the publisher's case with each renewal conversation.

Your Audience is More Powerful Than You Think

Publishers who combine a strong first-party foundation with third-party signal enrichment, lookalike modeling, and AI-assisted audience discovery are competing for a fundamentally larger pool of advertiser demand.

Optable brings all three expansion mechanisms and more together in one platform, with the agentic layer that makes them actionable at machine speed.

Schedule a demo to see firsthand how Optable’s AI-powered solutions can help you expand your addressable universe.

Data Management Platforms (DMPs) have helped publishers turn audience data into revenue for more than a decade. 

They solved an important challenge: collect signals, build segments, and make those audiences available to buyers. That model worked when digital advertising revolved around browser-based targeting, cookie-driven activation, and a relatively simple idea of what publisher data was supposed to do.

But those expectations have changed alongside the convenience and reliability of walled gardens and the mass adoption of AI tooling. What’s expected of publishers today is exponentially more than it used to be, and publishers need the right infrastructure to keep up.

Where Traditional DMPs Fall Short

Traditional publisher DMPs emerged to organize audience data and activate segments against demand. They collect data, associate users with signals, build segments, and pass those segments into ad servers, SSPs, DSPs, and other activation systems. For many publishers that workflow still applies, and DMPs perform well around web-based audience packaging and existing Ad Ops processes.

Audience creation is now only a part of what publishers manage. They are expected to curate niche audiences for buyers and to do it at machine speed across multiple platforms. The ecosystem they navigate has grown more complex:

• Identity across multiple frameworks and alternative IDs
• Omnichannel inventory spanning web, app, CTV, audio, and gaming
• Advertiser and retail media collaboration without exposing raw data
• Privacy-preserving data-sharing requirements
• Signal enrichment and addressability in the bidstream
• AI-assisted and agentic workflows for planning, activation, and measurement
• Faster RFP response cycles and more custom audience requests

Each of these depends on a foundation that understands how identity works across systems, partners, and channels. A DMP on its own does not meet these demands. Publishers need better identity infrastructure to satisfy buy-side expectations.

Don’t Treat Audience Tools as Identity Platforms

Identity platforms and audience tools are undeniably related, but they serve distinct functions. Audience platforms allow publishers to identify who to include in segments and where to activate their audiences. Identity platforms answer broader questions, such as:

• Who is this individual across channels and devices?
• How should different identifiers relate to each other?
• How do we collaborate with partners without exposing raw data?
• How do we support new activation paths as buyer requirements change?
• How do we give AI agents a trusted data layer to work from?

Most publisher monetization workflows depend on identity: Audience building, signal enrichment, clean room collaboration, omnichannel addressability, and AI-powered audience discovery all depend on a robust identity foundation.

Relying solely on a DMP for audience creation is becoming unsustainable. The demands for speed, accuracy, depth, value, and compatibility with agentic AI require publishers to re-evaluate their platforms and the foundation they’ve built upon.

The Publisher Workload Goes Beyond the Web

The old DMP world was built for a single-channel ecosystem. The modern publisher business is increasingly expanding to a cross-channel and omnichannel approach.

Web audience activation still matters, but it’s no longer enough.

A publisher may need to monetize authenticated readers on the web, anonymous visitors in Safari, app users, CTV viewers, podcast listeners, and gamers in environments where browser-based workflows do not work.

If each channel requires a different identity workflow, data model, and activation process, the publisher ends up dealing with more fragmentation. The result is more operational work, less consistent addressability, and a weaker first-party data story for buyers.

This is why identity needs to sit underneath the full monetization system. Publishers need a way to connect signals across surfaces, preserve the right level of context by user, household, and device, and activate that intelligence wherever the demand is flowing.

Time-To-Revenue Matters as Much as Audience Depth

DMPs were built for audiences that grow as users return and qualify. That can work for always-on behavioral segments, but it creates problems when the sales team needs to move quickly.

Consider a campaign where an advertiser sends a CRM file or a brief for a custom audience and wants to launch in two weeks. The publisher may have the right data, but if the live audience depends on visitors coming back to the site before they can be counted or activated, the campaign starts with less reach than it should.

RFPs move fast, campaign windows are short, and sales teams can’t wait for audiences to mature before they put out a proposal.

The stronger model is retrievable data from day-one: connect the data, match it to your existing identity graph, and gain enriched signals immediately ready for audience building and activation. Imported CRM data, partner data, and clean room outputs are ready for activation from day one.

Most DMPs Don't Do Signal Enrichment

A DMP helps publishers package and activate audiences, but identity infrastructure can directly improve the value of the bidstream itself through signal enrichment

When a publisher can resolve identity in real time and add addressable signals to a bid request, the publisher creates value before the impression reaches demand. That may include resolved IDs, inferred identifiers, audience membership, or other signals that improve addressability.

Signal enrichment gives the identity layer a direct line to monetization. It can be tested, measured, and tied to programmatic performance.

Most DMPs can’t fulfill this need. Signal enrichment requires server-side identity resolution, support for alternative IDs, and ways to improve coverage for unauthenticated traffic. It also needs to work inside the latency expectations of programmatic advertising.

Data Collaboration Needs Secure Identity Infrastructure 

Advertisers, agencies, retailers, and publishers increasingly need to work together without exposing raw customer data. Secure clean rooms make data collaboration possible, but they’re only as useful as the identity foundation underneath them.

A limited clean room workflow may help two parties match records and create an audience, but modern collaboration often requires more: activation and insights that both sides can trust, plus enrichment, measurement, and privacy controls that hold up under legal review.

The question is whether the platform can support the collaboration workflows publishers and advertisers actually use, across partners, identifiers, and governance requirements. Clean room platforms that connect to the same identity graph used for audience building, enrichment, and activation enables accurate data exchange and reduces the adtech tax incurred to manage separate tools for new partners and use cases.

AI Exposes Weak Data Infrastructure

A useful AI workflow for publishers needs access to the publisher's data, taxonomy, traits, audience definitions, identity graph, activation destinations, and business rules. It needs to understand what can actually be built, where it can be activated, and how it maps to the buyer's brief.

Layered on top of a weak identity foundation, AI tools will only generate inaccurate, incomplete, and irresponsible outputs. Whether that’s an audience that doesn’t actually exist or a deal that doesn’t align with budget, these unreliable results stem from lacking infrastructure. 

Closed AI tools inside a single vendor interface may help with basic tasks, but publishers and agencies are already standardizing on their own AI tools. The more durable approach is to make the publisher's data platform accessible to the AI clients and agentic standards teams already use.

An AI assistant that can suggest audience ideas is helpful. An AI agent like Optable’s Audience Agent that can understand campaign objectives, find relevant existing audiences, propose new ones, build them, price them, and push them toward activation is even more valuable. For AI tools across the spectrum, the right data foundation will make or break their performance.

What Publishers Should Look for Beyond the DMP

The next generation of publisher infrastructure can’t just recreate DMP workflows in a shinier interface. Updated infrastructure needs to expand and accommodate::

Publishers should look for:

• Identity resolution across web, app, CTV, audio, and other environments
• Support for user, household, device, and custom identity scopes
• Alternative ID support that is part of the platform, not a one-time activation signal
• Coverage for unauthenticated traffic
• Day-one use of imported and connected data
• Signal enrichment as a measurable revenue workflow
• Clean-room collaboration across multiple use cases
• Activation across ad servers, SSPs, DSPs, Prebid, walled gardens, and cloud destinations
• AI workflows that connect to the publisher's real data and preferred AI tools
• A commercial model that does not punish growth through unnecessary activation taxes

Publishers shouldn't just ask, "Which DMP has the best audience builder?" They should also ask, "Which platform can support the monetization workflows we need now, and the ones we will need next?"

The Identity Infrastructure for Agentic Audience Planning and Activation 

DMPs helped publishers create value from their audience data. The market has now moved beyond the problem DMPs were built to solve.

Publishers still need to find and build audiences, but they also need identity resolution, signal enrichment, clean rooms, omnichannel activation, and AI-ready workflows that adapt as new standards and buyer requirements emerge.

The next phase of publisher monetization will be built on identity infrastructure. Optable helps publishers build that foundation, so you can orchestrate audience creation, signal enrichment, data collaboration, cross-platform activation, and agentic planning from a single platform.

The DMP helped publishers monetize audience segments. Optable helps publishers monetize identity in the AI era. Schedule a demo today.

Prebid and Beeler.Tech hosted the second Trends, Tools, and Takeaways webinar, a quarterly session designed to help publishers get ahead of the shifts reshaping the open web. This quarter, the conversation circled the revenue opportunities starting to move through agentic advertising.

Mike Racic, President of Prebid and co-host, named it plainly, pointing to “a grand swell of revenue opportunities, particularly around the AdCP protocol, that most publishers should be paying attention to.”

The four sessions came at that opportunity from different sides but landed on the same point: agentic demand will reward publishers who have invested in the right foundations to receive it. 

Identity as the Foundation for Agentic Demand with Optable’s John Rosendahl

Optable's Director of Product Management, John Rosendahl kicked things off with some historical context: “We've always had an identity crisis, more or less since the dawn of programmatic.” In other words, the identity crisis isn’t new, but previous efforts to solve it haven’t stuck.

So, what’s different about the current moment? Publishers now run eight or more identity vendors at once, each with its own coverage, cost, and performance profile. Every ID a publisher carries quietly decides which buyers can reach it.

For the agentic era, that fragmentation stands to be a major constraint. When buyer and seller agents transact through interoperable standards, they act on the identity and audience signals they have access to. Insufficient or vendor-locked signals mean your inventory looks less appealing to the agents doing the buying.

John's answer was to treat identity as owned infrastructure: a flexible ID graph you control, rather than coverage rented from third parties whose terms and roadmaps shift underneath you.

He left publishers with three principles to act on:

1. Own your graph.
2. Build for flexibility.
3. Interoperate intentionally.

“Sustainable revenue is going to be built off identity solutions.” — Rob Beeler, Beeler.Tech

The Wider Conversation

Erik Svilich (Encypher) co-chairs the C2PA Text Provenance Task Force. He made the case for content authenticity that buyers and their agents can read.

Brands don’t want to run beside scraped pages, deepfakes, or AI slop, and machine-readable provenance lets original reporting prove what it is. That signal only grows in value as agents decide where to spend.

Brian Sardo (Microsoft) walked through a model for Prebid Server as a publisher-controlled decision engine, coordinating demand while leaving delivery to the ad servers publishers already run.

Kieran Greene (Shinka) and Karim Mourra (JWX) closed on video caching and migration, where the goal sounds simple but is harder to deliver: cut latency without giving up control.

What This Means for Publishers

The agentic opportunity Mike opened with is beginning to move, and the sessions made it clear that publishers can’t capture it by signing up with another vendor. They need to own the foundations agents will read, with provable content, identity and audience signals they control, and decisioning that stays in their hands.

For most publishers, the first move is the one John outlined: Own your identity graph, keep it flexible, and make it interoperable. When agentic demand goes looking for your inventory, it can find the signals you own and can stand behind.

Optable is the agentic audience platform building the identity, audience, and activation infrastructure for publishers. To see how that holds up against the fragmentation John described, book a demo. And to go deeper on the agentic shift driving this quarter's conversation, download our guide here.

In a live poll during Optable’s Beyond the Hype breakout, 51% of attendees described themselves as cautiously optimistic about agentic advertising. Twenty-eight percent identified as sky-high believers, compared to six percent who called themselves outright skeptics.

That mix was the spirit of the Summit. We all felt the urgency, supported by questions about whether the infrastructure, standards, and playbooks will catch up to the ambition.

The day opened with Tim Berners-Lee and closed with WPP Media. In between, more than 20 sessions argued one point from different angles: the agentic web must be built on shared standards, or it will be built badly. The room mostly agreed, but disagreed on where value lands in the agentic stack, what happens to publisher business models in transition, and whether operators can really control systems designed to act on their behalf.

Here are the five threads that ran through the day, what speakers said about each, and where the harder questions still sit unanswered.

‍

1. Standards as the load-bearing concept

Anthony Katsur framed the day with a single arc: from Turing to the agentic web, the throughline is open infrastructure. Extending programmatic’s philosophy on standards to the agentic layer is the natural move.

The named pieces stacking up:

• AAMP as the agentic protocol layer on top of OpenRTB.
• AdCP as the open standard for agent-to-agent communication, positioned as parallel to OpenRTB rather than a replacement.
• MCP as the standard for agent-to-system context.
• CoMP as the standard for metered, accountable AI content access.
• Cloud deployment patterns for shipping these standards into production environments at scale.

Whether the protocols ship before the platforms route around them is the remaining question.

2. Where in the agentic stack does value land?

This was the most contested thread of the day, and the most useful one for operators trying to plan a roadmap. 

Three positions on the same stack:

• The buy-side orthodoxy. The Trade Desk's Eric Bosco and Stagwell's Dru Sil described agentic AI aligned to outcomes, built on high-quality inputs, always under your control.
• The contrarian view. KERV.ai's Marika Roque pushed back. Bidder optimization, SSP tuning, ad serving, and DCO already run in real time. The opportunity for agentic systems sits in the inputs: planning, audience construction, and the brief itself. Roque argues that 15 years of programmatic optimization have already harvested most of the gains available inside the auction.
• The cross-stack view. The “Agents Across the Ad Stack” panel, moderated by Optable's Bosko Milekic, made the case that the value looks different at every layer. Mike Wolk (Goodway Group), Nishant Khatri (PubMatic), and Scott Menzer (Unity) weighed in. Goodway is rolling out agentic audience workflows to its buyers. PubMatic is shipping agent-to-agent transaction infrastructure. Unity is putting agents on top of runtime behavioral data from mobile gameplay.

None of these are wrong, but they’re not complete on their own, either. Wolk's framing captures the prerequisite the rest of the day mostly assumed: AI on top of fast, wrong data is still wrong. Meanwhile, demand for agentic media is being shaped by buyers, with most of the supply side still watching from the sidelines.

Keep reading: Get Ready for AI Agents Now, Before Buyer Agents Scale

‍

3. The publisher business model is transitioning in public

Drawing on a network of news organizations, INMA's Gabriel Dorosz painted a picture of traffic disruption from AI-mediated discovery and a business model transition still in progress.

What the transition looks like in practice:

• New monetization primitives. Supertab's CoMP turns AI crawling into a settled commercial transaction, with metering, pricing, and accounting designed for an environment where the page view is no longer the unit of value.
• Premium supply, agentic access. The framing of a publisher operating system that lets buyer agents traverse OpenRTB and AAMP is the most concrete answer in the room to what replaces traffic, clicks, and pages.
• The platform view. Jessica Chan's Perplexity session framed the web as moving from clicks and queries to intelligent discovery.

4. Identity and consent assumptions break when agents act

The identity panel with Will Oatley (Zeroma), Adam Pyett (Ogury), Mathieu Roche (ID5), and Rowena Lam (IAB Tech Lab) took on the question of how identity frameworks need to evolve when an AI agent is the one browsing, searching, and transacting.

Today's identity infrastructure was built to connect a human across devices and channels. The agentic era introduces a new layer on top of that: an agent acting on a user's behalf, with its own session, context, and path to purchase.

The two surfaces under pressure:

• Identity. When an agent acts on behalf of a user, whose identity attaches to the resulting impression? Which framework reconciles AI, online, and offline signals into a coherent path to purchase?
• Consent. The legal architecture of digital privacy assumes a human reads a banner and chooses, but agents do not read banners. The consent surface the industry spent a decade building risks losing its addressee.

LiveRamp's Christian Carlsson framed agentic audiences under AAMP as the reconciliation layer, with transparency and control as the design principles. Whether transparency to a human and transparency to an agent require the same thing is an open question.

5. The control paradox

Operators want control, but the emerging technologies stand to remove it.

Two positions in tension:

• Control as a feature. The Trade Desk and Stagwell sold their session on agents being “always under your control.” Snowflake's activation framing centered the same value. Meta's Artur Souza described the move from scripted responses to context-aware, action-oriented agents, which is, read plainly, a description of moving the human further out of the loop.
• Control as illusion. Mobian's Jonah Goodhart pushed back. In an AI-first world, brands can’t fully control how they appear. Context determines meaning, and the brand is not the author of its own context.

What happens when agents transact with other agents and the human is one step removed from each decision?

The honest answer is that operators can retain control over inputs, objectives, and guardrails, but they may eventually surrender the supervision of moment-to-moment decisions agents make. That trade is the deal of the agentic era.

What didn't get said

There were a few absences worth flagging, because they are the topics the industry will have to address.

• Measurement. How do agent-mediated impressions get measured against agent-mediated outcomes?
• Buy-side economics. When an agency runs an agent on behalf of a brand, who pays for the compute, on what model, and how does that cost compare to the human equivalent it displaces?
• The non-US view. INMA brought a global lens, but the European regulatory reality remained relatively unaddressed.
• Where the hype has outpaced reality. The honest answers about which agent deployments stalled or never delivered against the pitch weren’t discussed, leaving little room for collaborative problem solving to overcome the challenges.

What we took away

Agreeing on standards is an essential step. The harder work, which will decide whether the agentic web pays operators back, lies in audience construction, identity reconciliation across human and agent signals, and commercial primitives like CoMP and AAMP that turn agent activity into accountable transactions.

For the industry, the urgency centers on AdCP, AAMP, MCP, and CoMP shipping in time to matter. For organizations ready to move forward, the work lies in auditing their agent readiness, identifying where agents create real lift today, and building the infrastructure that makes agent-to-agent collaboration possible when it arrives.

Where Optable sat in the conversation

Bosko Milekic moderated “Agents Across the Ad Stack” with Mike Wolk (Goodway Group), Scott Menzer (Unity), and Nishant Khatri (PubMatic). This session discussed where agentic AI is currently creating real value for the buy side, sell side, and platforms. 

John Rosendahl and Andy Sharkey led the “Beyond the Hype” breakout on what agents actually do for media buying and selling workflows today. The session was built around the Optable & Digiday State of Agentic Advertising survey, live audience polls on agentic AI, and a demo of Optable's Audience Agent. The structural argument: the hype curve looks different once the technology is sitting in front of you. Audience Agent sessions grew 89% from Q1 to Q2 with the customer count doubling, and Q2 is still in progress.

Both sessions returned to the same question: what actually changes on Monday morning? Audience Agents in the RFP workflow and Sales Agents representing publishers in the ecosystem are how Optable is answering it, creating the foundation for the Agentic Collaboration Marketplace.

If you were at the Summit, we'd love to compare notes. If you missed it, we’re continuing the conversation at Cannes Lion in June.

AI agents are already reshaping how advertising is bought. Advertisers are deploying them to plan campaigns, evaluate publisher inventory, and negotiate deals—at machine speed and at a scale no human team can match.

This shift is happening inside today’s RFP and direct-sales process, not in some future marketplace. Buyer agents are increasingly determining which publishers are considered, how their audiences are valued, and how quickly deals move. For publishers, the implication is immediate: if your inventory cannot be understood and evaluated by these systems, it risks being overlooked entirely.

Agent readiness is the work of ensuring that doesn’t happen. It is the process of preparing publisher data, audiences, and inventory to be discoverable, interpretable, and competitive in an environment where machines are the first point of evaluation.

There are two reasons to prioritize this work now rather than waiting for the agentic marketplace to mature:

1. The return on agent-ready infrastructure shows up in the current RFP cycle, not just in some future quarter when buyer agents become the standard.
2. The standards governing agentic advertising are being written right now, and publishers who wait will inherit rules that others have already defined.

See a shift by the next RFP cycle

The work that makes a publisher’s inventory available to a buyer agent has immediate value inside the current direct-sales motion. Agent readiness depends on a handful of foundational assets:

• A unified identity graph
• Audience segments enriched with verified third-party attributes & your own structured first party data
• Inventory packaging that answers what’s available, who you reach, and at what price

Each asset strengthens your next RFP response while preparing your inventory for agentic evaluation. Buyer agents are already querying publisher inventory at machine speed, and sales agents that can directly respond to them are emerging. When that agent-to-agent communication matures, deal cycles condense from weeks of human back-and-forth into minutes of structured exchange between systems.

Signal and audience agents are already running on the publisher side, synthesizing and packaging audience data into outputs that sales and buyer agents can act on.

The payoff will show up within the quarter, long before the agentic marketplace fully develops.

Shape the standards being written today

Agentic advertising only works at scale if buyers and sellers share a common language for agents to discover audiences, negotiate deals, exchange brand safety signals, and enforce publisher-defined rules. Built on MCP, the Ad Context Protocol (AdCP) is defining how agents transact across the advertising ecosystem, and the IAB Tech Lab is running a similar agentic track with their AAMP framework.

What matters to publishers is that open, interoperable standards prevail, and that publisher interests shape the outcome. An interoperable standard gives your premium audiences a path to buyers beyond walled gardens. It lets a buyer agent discover, evaluate, and transact on the value of your inventory directly, at prices that reflect the value of your audience, without routing that value through a closed ecosystem or requiring manual negotiation at every step.

The working groups developing these protocols are open, and they need publisher voices to weigh in. A few ways to engage:

• Join the groups shaping the standards. Organizations like the AgenticAdvertising.org and  IAB Tech Lab run open agentic tracks. Membership gets you a seat in the conversations where standards are being defined.
• Assign a data or revenue strategy operator to attend. You don’t need a senior engineer, just someone who understands your inventory well enough to read proposals and respond.
• Submit comments on the decisions that affect your business. Share your opinions on inventory representation, audience description, pricing, brand safety, and consent.

How to become agent-ready

Buyer agents are moving faster than publisher workflows can keep up with. Three operational priorities close that gap:

• Assess honestly. Run a cross-functional audit across data engineering, ad operations, revenue, and legal to map where you stand across the pillars of agent readiness: data, inventory signals, content and context signals, technical infrastructure, and organizational awareness.
• Pilot on a real workflow. Pick one property, audience, or deal type. Configure an agent against that scope, establish a control group, and run it against live direct-sales RFPs over a defined window. Document performance and evaluate how you can move forward.
• Continue shaping the standards. Keep participating in these governance efforts as you run your pilot. The decisions being made in parallel to your work will define the marketplace that emerges.

Agent readiness is the prerequisite for the future of AI-powered media selling and agentic collaboration, and investing in it now benefits your manual and programmatic workflows today.

Start the Audit, Run the Pilot, and Claim Your Sea

Agent readiness is operational work that pays off inside your current workflow and positions your organization for the emerging agentic marketplace. The publishers who start the work now will enter that marketplace with a foundation their competitors are still scrambling to build.

‍

___

This article is originally published on Digital Content Next on May 4, 2026.

CTV ad spend is growing, but most publishers aren't capturing their share. Without a robust identity infrastructure, publishers can’t monetize their rich audience data at a premium price, connect their audiences to valuable signals or transaction data, or transact at the speed the market now demands.

Two Distinct Identity Challenges

CTV poses two unique challenges: siloed data and unauthenticated data. Without a clear idea of who their users are, CTV platforms can’t accurately resolve identity, create addressable audiences, or implement agentic workflows.

For premium streaming publishers, broadcaster apps, and subscription video services, rich data is available. Viewers have logged in with their own accounts, but this login data sits in separate systems that don't talk to each other. A major media company might have authenticated viewer data in a news app, TV streaming app, and magazine subscriptions with no connection between them. Advertisers can't reach the same person across properties because the publisher's systems don't recognize it's the same person.

For free, ad-supported streaming (FAST) publishers, the situation is different. No login is required means no email address and no account data. With device ID or household IP, identity resolution becomes more difficult. A viewer watching a free channel on Roku or a Samsung TV is effectively anonymous to the ad system, even if they sit in a household with a known address, purchasing history, and browsing behavior on other devices.

Both problems cost publishers money. The solutions are related but distinct, and an identity infrastructure worth building needs to handle both.

Why the Web Playbook Doesn't Translate to CTV

Publishers who approach CTV with web-based tooling quickly realize that the methods that work in a browser don't function in a streaming environment.

The tag-based approaches that most web identity infrastructure uses don't apply. Identity resolution on CTV has to happen server-side, which is a meaningfully higher technical lift than adding a script to a web page.

The identifier landscape compounds this. The signals that exist across streaming platforms come in different forms and don't naturally connect: Roku IDs, Samsung TV device identifiers, Fire TV device IDs, household IP addresses, and hashed email addresses all live in separate environments.

The household IP is the most useful signal available for FAST audiences. A household that watches free content and visits the publisher's web properties can be linked via residential IP. Adding web behavioral data makes the inferred CTV profile much stronger.

Solving CTV identity in isolation tends to undervalue what publishers already have access to. A publisher whose CTV data lives separately from their web and mobile data surrenders targeting precision and remains invisible to the automated systems buyers are already using to discover and activate against publisher inventory.

‍

‍

What CTV Identity Infrastructure Needs

The underlying infrastructure needed is the same regardless of which problem a publisher is trying to address.

Both require a server-side resolution layer that evaluates available identifiers for a given impression and selects the strongest one in real time. That selection happens automatically, with consent and privacy enforcement built into the workflow. Publishers can then directly connect to CTV-specific ad servers and SSPs including Publica, Freewheel, and Magnite.

Publishers with siloed authenticated data need to resolve the same viewer's identity across sources, including:

• TV streaming and VOD apps
• News apps
• Magazine  subscriptions
• Web properties

Publishers with unauthenticated inventory need to make inferences about their users to get a clearer idea of who they are. The path is straightforward:

1. Collect Household IP addresses and device identifiers from the streaming environment
2. Map these identifies to email addresses through data partnerships
3. Translate them into alternative IDs the programmatic ecosystem recognizes
4. Share the resolved identifiers to the ad server without ever needing an email

The same identity foundation should handle both resolution paths. Agentic audience workflows require a single, queryable graph to function. A fragmented infrastructure can only ever propose partial audiences, and partial audiences don't win briefs.

The Agentic Payoff

Identity resolution is the prerequisite for unlocking agentic revenue opportunities. The payoff becomes possible when publishers don’t have to rely on manual audience packaging and activation. Agentic workflows enable efficiencies and opportunities, including:

Local Household Targeting

Connecting household-level identity to geographic and behavioral signals creates high-value inventory packages for local and regional advertisers. But the audiences need to be continuously refreshed as new viewing data comes in, and manually matching them to incoming briefs at scale isn't viable.

Optable’s Audience Agent reads an incoming brief, queries the identity graph for matching households, builds the audience, and routes activation through integrated SSPs. A workflow that previously required a data analyst and a campaign manager now completes in minutes.

Content Alignment

Publishers with strong editorial signals can command a premium by matching advertiser brand requirements to specific content environments. Performing that analysis across a large CTV catalog and keeping it current requires automation.

Optable’s lightweight SDK continuously collects and organizes your 1P data, giving the Audience Agent up-to-date information for audience discovery, packaging, and activation.

Addressable FAST Inventory

FAST publishers have large impression volumes and low baseline addressability, meaning even modest resolution rates produce meaningful revenue. Because those audiences are built from inferred signals rather than deterministic data, they need to be rebuilt continuously as viewing behavior changes.

Optable's Audience Agent uses content context and behavioral signals to construct audience packages that give buyers a credible reason to pay a premium on inventory they might otherwise overlook.

Across all three applications, the practical effect is the same: publishers with lean teams can handle more volume, respond to briefs more quickly, and offer more customized proposals without adding headcount.

CTV Revenue Begins With Identity

Publishers with siloed authenticated data need to connect what they already have. FAST publishers need to build from the signals they have available. These starting points converge on the same outcome: a unified identity and audience infrastructure that spans every environment where viewers show up, in a format that buyers and buyer agents can act on.

Publishers need a robust identity infrastructure to build addressable audiences, manage ad frequency across devices, close the measurement loop, and participate meaningfully in the emerging agentic ecosystem. The identity work done today determines which publisher inventory is visible to buyer systems tomorrow. 

Ready to build an identity foundation for the agentic era? Schedule a demo today.

More RFPs mean more deal opportunities, but for leaner Ad Ops teams, meeting advertisers’ rising expectations can be increasingly difficult. Advertisers are increasingly looking for personalized, unique approaches to their RFPs.

Manually responding to an RFP takes too long to keep up with those time-sensitive demands, especially when advertisers are using buyer agents to query your inventory.

The Cost of a Slow RFP Response

Unanswered RFPs leave money on the table. The manual response process doesn’t meet the market’s demands for volume, speed, and specificity.

The impact of slow responses extends beyond fumbling isolated deals:

• Lean Ad Ops teams are faced with higher volumes of incoming RFPs, increasing the gap between demand and bandwidth.
• Publishers responding under pressure tend to default to familiar, broad segments, resulting in undifferentiated proposals to buyers who expect precision.
• Teams that are at capacity may sideline incoming RFPs, which registers only as unanswered briefs instead of lost revenue.
• Advertisers may deprioritize publishers who don’t respond with audience recommendations at the pace and depth they’ve come to expect from walled gardens.
• As more advertisers deploy buyer agents to search for inventory, a manual workflow has no answer for a buyer agent querying your inventory outside business hours.

The solution isn’t just to speed up the process but to improve the specificity of audience proposals. Agentic workflows supported by platforms like Optable make it possible to surface audiences that manual processes routinely miss and close the loop from RFP to activation without the bottlenecks.

How Agentic Workflows Are Changing the Sales Process

No matter how good the team is, a manual RFP workflow limits how many briefs they can strategically answer in a day. Agentic workflows remove that cap by changing what happens at each stage of the sales process.

When a brief arrives, instead of routing it to a queue and waiting for a data analyst to pull segments, the agentic workflow begins immediately. An agent can read the brief, extract campaign goals, and generate audience recommendations using the publisher’s enriched first-party data. Publishers with low or declining traffic can more effectively monetize their audiences when an agent surfaces precise, differentiated segments to target.

This process surfaces segments that a manual query would overlook: niche behavioral combinations, trait correlations, and audience overlaps that only become visible when systematically searching the full depth of the data. Optable's Audience Agent handles this step, proposing ranked segments with confidence scores and reach estimates that give the sales team something to review rather than build from scratch.

The team can adjust in natural language to instantly update the recommendations. A human reviews before anything activates, keeping the publisher in control of what goes out and to whom. Then, the agent can generate a response draft and activate audiences directly across ad servers and SSPs. The proposal, complete with audience definitions, reach estimates, and pricing, is ready to send in a fraction of the time.

Agentic workflows also change how buyers and sellers interact. As agencies and brands increasingly deploy buyer agents to search for inventory, publishers who aren't visible to those systems are out of consideration. A Sales Agent like Optable’s allows publishers to define their audience products, configure pricing and access rules, and make that inventory automatically discoverable and transactable to buyer agents querying for relevant audiences.

Publishers with premium audiences get access to more deals, and buyers receive a structured proposal mapped to their brief.

What Your Team Can Do With the Time Back

When proposal building stops eating up your team’s bandwidth, the work that requires human judgment gets the attention it deserves. For Ad Ops teams, that looks like:

• More thorough campaign analysis: With a tool like Optable’s Audience Agent handling brief intake and segment recommendations, teams can focus on surfacing compelling insights around their advertiser’ audience to build smarter packages around performance instead of availability.
• More niche, addressable audience products: With agentic support, teams can develop more compelling audience products that go beyond standard segments without allocating more resources to surface them.
• Unique, custom creative approaches: Brands aren’t looking for cookie cutter proposals. They want to see platforms with custom audiences built from unique first-party data that aligns with their vision, and agentic solutions give Ad Ops teams the time and ability to deliver.
• The ability to answer more briefs: Lean teams can address briefs that they previously would have deprioritized or overlooked due to bandwidth constraints, increasing revenue potential without expanding headcount.
• Stronger advertiser relationships: Teams that aren’t weighed down with manual processes have more time for data collaboration with buyers. Sharing audience insights, validating campaign fit, and collaborating on strategy builds transparency that fosters a long-term partnership.

Agents help relieve the mechanical burden of processing and responding to RFPs so your team can focus on more meaningful tasks that require creativity, strategy, and human judgment.

Win More Deals With Smarter Workflows

Ad Ops teams with the right infrastructure to support them can respond faster, propose better audiences, and build stronger advertiser relationships without working harder.

Agentic workflows change what a lean Ad Ops team can accomplish in a day: more brief responses, more precise audiences, and more time for the strategic work that turns one campaign into a long-term relationship.

Every brief the agent handles well is data that informs the next one. Every niche audience that wins a deal becomes part of a stronger audience product portfolio. 

The buying landscape is moving toward greater automation on all sides. Publishers building agentic infrastructure are positioned to compete in that environment and won’t have to scramble to catch up.

Optable gives publisher teams the agentic infrastructure to do exactly that, from RFP intake and audience activation to outbound discoverability. Ready to see it in action? Schedule a demo.

We returned to the Beeler.Tech Navigator event this year to join fellow industry leaders in conversations about latent data, signal quality, practical AI applications, and more. Optable’s Director of Sales, Kristy Schafer, led a session with Tim Hurd, VP of Media at Goodway Group, to share how the agency rebuilt their audience stack.

That initiative produced Passport One, Goodway’s proprietary data and intelligence layer that centralizes audience creation and decisioning with the ability to activate across the ecosystem. Optable provides the identity, audience, and collaboration layer where clients onboard first-party data and Goodway centralizes its enterprise data partnerships.

‍

‍

What Goodway Built

The motivation was straightforward. Goodway wanted a system that allowed them to centralize audience creation and decisioning with the ability to activate across the ecosystem.

In practice, that resulted in:

• Client data onboarding dropped from four weeks to roughly five minutes.
• The data fees stacked into every activation came down.
• More spend now routes directly to the source: the publishers themselves.
• Planners can index and overlap consumer segments to hit multiple consumer profiles with more strategic impression targeting.

The goal throughout: more value out of every working dollar, with more of that revenue passing through to the content owner.

Why They Built It

Tim spent meaningful time on the why behind the work. Goodway's clients aren't simple. A single brand often needs to reach multiple consumer personas at once, against very different media outcomes across awareness, in-store visits, and new account creation.

That kind of complexity demands a centralized audience layer, where the same audience can be pushed out to a multitude of destinations rather than rebuilt for each one. His point landed cleanly: the more easily publishers can plug into that workflow, the easier it becomes to spend with them.

What This Means for Publishers

Asked what he most wants from publishers going forward, Tim was direct. He hopes publishers will build similar infrastructure on their side in the form of interoperable audience platforms.

The next phase on this journey is agent readiness. As both buyer- and seller-side agents evolve, publishers with a strong identity and audience layer will be the ones surfacing inside the agency's buying strategy.

Agencies will gain the ability to build audience strategies across publishers using the underlying metadata of each one's identity and audience infrastructure. In turn publishers can better monetize their first-party data while still allowing the agency to buy at scale from one centralized location.

What Happened to CPMs?

The question that probably mattered most to the audience came from a publisher in the room: what's happened to CPMs through all of this?

Tim's answer: CPMs have come down. The agency gets the efficiency it set out to capture, and more of that working media is now flowing straight to the publisher instead of being absorbed by the middle.

The broader upside for publishers willing to evolve in this direction:

• Less ad-tech in the way
• More paths to buying
• Broader inclusion in plans

That dynamic is exactly the upside Tim's pitch holds out for publishers willing to evolve in the direction he's describing.

He closed with a line that captured the whole posture:

‍"Democratize data access, and look at each other less as competitors and more as a cohort."

‍