AI is already changing how media gets bought: ads are appearing inside chat interfaces, open-web traffic is shifting toward premium publishers, and buyers are starting to hand campaign work to software agents. This session brought an analyst's view to a day of practitioners, using EMARKETER's forecasts to size what's coming.

Yory opened with how AI advertising will benefit, and even create, walled gardens. As an extension of direct demand, agentic advertising can level the playing field for smaller players. Agents can also make programmatic itself work better, with richer signals, faster decisioning, and more transparency, and agents can negotiate time-sensitive and real-time inventory that's hard to sell today.

On standards, Yory contrasted two approaches. AAMP began as an evolutionary path by bringing agents into programmatic and media planning. AdCP began as a revolutionary approach, with a reimagined workflow across the whole ad stack. Because the two increasingly overlap, he argued, the bigger priority is interoperability, not which set of standards wins.

He projected a large increase in agentic ad spend by 2028, while noting that behavior changes slowly. His four steps to prepare: get specific about what you want from agentic in your goals and briefs, set clear governance rules, organize your data so agents can use it, and experiment.

Digital Content Next (November, 2026) — In 2020, DCN reported on a study that put a number on what many publishers had long suspected. ISBA and PwC traced programmatic spend from advertiser to publisher and found that publishers received just 51% of it. Despite a push for transparency from both sides, the structure hasn’t improved much over time. An open-market buy still passes from agency to DSP to exchange to SSP, with data, identity, and verification vendors along the way. Every hop is an additional cost, taken from the publisher’s share.

Now, the industry is building its first new way to transact since programmatic. Agent-to-agent buying could shorten the distance between buyer and publisher, or it could add a new toll at every step. Which one we get depends on the terms being set right now, while the market is still small enough to shape.

In agentic advertising, we can return to a simpler standard: a seller, a buyer, and a direct connection between them.

How we got here

Programmatic solved many problems in media buying. SSPs let publishers sell supply at scale, and DSPs gave buyers one place to manage spend. Every layer that followed, for data, identity, verification, and measurement, fixed something else.

Individually, these solutions added value. Stacked together, those solutions became a toll road, with every vendor between buyer and seller taking its share.

Direct deals at programmatic speed

Direct deals have always depended on conversations: buyers explaining what their campaign needs, and publishers explaining what their audience and content can deliver. This approach leads to strong campaigns, but it’s also why direct deals take so long to finalize.

For most campaigns, a week or more of emails, RFP responses, and manual planning isn’t worth the effort, so the budget goes to the open exchange instead.

Agentic buying cuts the back-and-forth down to minutes while keeping the buyer and seller in direct contact. Here’s how an agentic direct deal works:

1. Share the brief. The buyer’s agent sends a campaign brief to a publisher’s seller agent, outlining the buyer’s objectives and requirements.
2. Discover relevant ad products. The seller agent matches the brief to the publisher’s ad products—combining inventory, data, and pricing—and returns one or more options for purchase.
3. Select the buy. The buyer chooses which ad products to purchase (a human-in-the-loop activity today), and the buyer’s agent communicates that selection to the seller agent.
4. Review and approve. The proposed order goes to the publisher’s team for human review. The team evaluates the terms and approves the buy if they are acceptable.
5. Book the order. Once approved, the seller agent books the order into the publisher’s own ad server.

What comes out is a direct IO that moves at programmatic speed, with fewer intermediaries.

Or the same tolls at machine speed

None of that is guaranteed. Nearly every intermediary in the programmatic chain is building an agent of its own. If each one sits between buyer and seller, the deal picks up more fees, collected even faster than before.

The new tolls just look a little different:

• A take rate on what the publisher is paid
• A charge for every brief answered or task completed
• Activation fees to put audiences or inventory in front of buyer agents
• Clearing fees deducted from the publisher’s payout

Together, they rebuild the same stack of fees inside a new interface.

The first platforms to scale will set the defaults for how agentic deals are executed and priced. Publishers have seen this before: programmatic’s fee structure was standard long before they had the leverage to question it.

What fair agentic economics should look like

Standards bodies like AgenticAdvertising.org and IAB Tech Lab are working out how agents should talk to each other. How the money moves is up to the market. Publishers can help set those terms by holding platforms to principles such as:

The deal stays direct. Any platform that places itself between them should be able to say exactly what it adds to the equation and justify what it charges.

Platforms charge for software, not a share of the deal. A marketplace that takes a percentage of every transaction earns more as deals get bigger, whether or not it does more work. Pricing the platform as a service keeps the deal value with the people who create it.

The agent answers to the publisher. An agent paid on spend or take rate works for its operator. A publisher’s agent should carry the publisher’s prices and rules, and nothing should go live until a person approves it. When we surveyed 105 attendees at Optable Connect, two-thirds said they want agents to negotiate and prepare orders, with people approving the final buy.

The buy-side won’t wait

Publishers aren’t the only ones who want better, more direct deals. Independent agencies like Goodway Group are prioritizing direct publisher connections and trimming SSPs as part of supply-path optimization. Some buyers are already deploying buyer agents to transact on their behalf.

What that means is that buyers aren’t waiting for publishers to see the money first. Publishers who build now will help shape how agentic deals get done.

A direct model protects what this business has always run on: relationships. When buyer agents and seller agents talk, buyers and sellers are talking, too.

The next era of advertising

The value of agentic advertising comes from removing the distance between buyer and publisher. That’s why we launched the Optable Agentic Marketplace in October, with a 0% take on direct deals between buyers and Optable seller customers, that provides buyers a single connection point to connect to multiple seller agents.

For publishers, the first step is small. Pick one high-value package, price it, and put it in front of a buyer agent. Then look closely at how much of the buyer’s spend reaches your ad server compared to an equivalent buy through another channel.

The new standard should be the one the industry started with: direct connections between buyers and sellers, now possible at scale through their agents, with fewer intermediaries and better data between them.

Earlier this year, Optable ran the State of Agentic Advertising survey with Digiday and found that more than half of publishers cite internal readiness as a barrier to agent-driven demand, second only to technical complexity.

The technology itself gets most of the airtime in agent readiness conversations. Yet an agentic deal moves through Ad Ops, Revenue, Data Engineering, Legal and Product, and it stalls wherever ownership is unclear. We created the Agent Readiness Self-Assessment & 90-Day Plan to help teams of all sizes identify where they stand today, with a phased plan to close the gaps.

The Roadblock is Coming From Inside the Organization

Agentic deals can run as direct orders or programmatic deals, so they depend on teams that usually work on separate tracks. A single request from a buyer agent can touch packaging, pricing, audience data, and consent at once. Buyer agents evaluate multiple publishers at the same time and may commit to the fastest credible option. A request that waits on a manual handoff between teams risks going to another publisher.

That makes response speed an organizational question as much as a technical one. The primary hurdles to overcome:

• Knowledge and ownership sit with too few people. Unprepared teams may only have one or two people who understand agentic AI and no shared ownership. Every handoff between Sales, Ad Ops, and Data runs through those few people, and work stalls when they’re busy or away.
• Legal questions remain unanswered. In the State of Agentic Advertising survey, 37% of publishers cited regulatory or compliance uncertainty as an obstacle to agentic adoption. Without a clear answer on consent for agentic data exchange, deals get stuck.
• Training hasn't been funded. The same report concludes publisher teams likely need more education to win internal buy-in and resources. Respondents ranked education and organizational change among the least-invested areas.
• Bandwidth is stretched. Many publishers are rebuilding data infrastructure while core revenue is under pressure, and there’s only so many hours in the week. Adding headcount to tackle the project may not be an option, especially for smaller publishers.

Without training, knowledge stays concentrated. Without clear owners, a stretched team can't decide which hurdle to clear first. Deciding on ownership is where your organization needs to start.

No Team Gets To Sit This One Out

An organization that isn't ready only has one or two people who understand agentic AI, no shared ownership, and workflows that are mostly — if not fully — manual. In an agent-ready organization, every team knows its role, responsibilities are defined, and Sales and Ad Ops use agentic workflows as part of their routine.

Here's what readiness asks of each team:

None of these roles work in isolation. Readiness requires a cross-functional team with clear ownership, responsibilities, and mutual understanding.

Let Agents Tackle the Legwork

When agents take over research and setup, people can prioritize strategy, judgment, and approval.

The Weather Company shows how that split works in practice. Its agents handle audience discovery, linking real-time weather conditions to consumer behavior, while the team provides final verification for accuracy and brand safety. Benny Pang, product leader for audience data, calls the agents “a force multiplier.”

Turnaround times shrink as well. At Connect, Goodway Group described going from a client brief to audience segments in under an hour, compared to days or weeks when tackled manually.

For publisher teams, the change shows up in where the hours go. Ad Ops spends less time on manual setup and more time on the deals that perform. Sales packages audiences faster and answers RFPs with more precision. The time that’s freed up gets allocated to direct deals, new audience packages, and partner relationships.

People stay in the loop. Two-thirds of Connect respondents want agents to have more room to negotiate and prepare orders, with humans getting the final say. Where exactly that line sits is a decision each publisher will need to make.

From Kickoff to Agent-Ready in 90 Days

Ninety days is enough time to get to a working pilot with named owners, early results, and agreed rules. Jumping straight to adopting new technology tends to backfire. Without a shared foundation, an agent can end up pitching audiences nobody vetted at prices nobody approved.

The plan below runs in three phases. Each one names who leads, what to do, and how you'll know you're ready for the next phase.

Day 1 to 30: Assess & Organize

Data Strategy and Product lead this phase, with Ad Ops, Revenue, Legal, and Data Engineering playing their parts. As part of this phase, be sure to:

• Stand up a cross-functional working group and give it decision rights over the readiness plan. Book a recurring meeting before the first one ends.
• Record where you’re starting. Document authenticated traffic, addressability, enrichment coverage, and how long RFP responses take today.
• Identify and rank your organization’s gaps by impact. Assign someone to each one.
• Map your current integrations against the agentic frameworks your partners support.

You’re ready to progress when every gap has an owner, you’re written down your baselines, and the group is meeting on a regular schedule.

Day 31 to 60: Build & Activate

Ad Ops and Data Engineering take ownership during this phase, with Revenue shaping what gets tested and Legal reviewing. During the second phase, prioritize the following:

• Choose one pilot segment amounting to a single property, audience, or deal type that Revenue believes buyers want.
• Agree on scope, timeline, and success criteria so every team measures the same thing.
• Record content metadata and taxonomy; clean it up for the pilot inventory, and document the changes.
• Work with at least one partner that can activate both direct orders and programmatic deals, then set up your agent to surface the pilot audiences.
• Maintain a control group from day one. Compare results against your baselines each week.

You’re ready to move onto phase three when your pilot is live, the control group is clean, and you have early numbers to review.

Day 61 to 90: Sell Agentically

Revenue Strategy and Data lead the third phase, tapping Ad Ops and Legal for support. During this phase, you’ll need to:

• Decide how your seller agent represents you. That might include which audiences it surfaces, how it describes inventory quality, which deal types it puts first, and how it answers buyer questions.
• Write governance policies for security, accuracy, transparency and consent, and push them through legal review.
• Report agentic-sourced revenue as its own line so leadership can see what the pilot produced.
• Train each role on its part of the workflow. Practical use cases might include setting guardrails, validating outputs, and knowing where human judgment still has a role.
• Pick the next segment to bring into the program.

You can start scaling when you’ve configured your seller agent, policies are approved, agentic revenue shows up in reporting, and each team has been trained on its role.

Start By Getting Everyone in the Same Room

An agent-ready organization doesn't need every answer on day one, but it does need a plan to educate and empower your people before you can start transacting agentically. The technology matters, but it performs only as well as the teams running it.

Buyer agents are already comparing publishers on how fast and how clearly they respond. The publishers those agents return to will be the ones whose teams can speak for their inventory without a week of internal handoffs.

Optable Audience Agent gives Sales and Ad Ops turns a client brief into audience segments your team reviews. When you're ready to sell, Optable Sales Agent represents your inventory to buyer agents, and the Agentic Marketplace lets them place orders directly while you keep control of pricing and approvals.

If you're working out where your teams should start, schedule a meeting and we can walk through it with you.

Programmatic proved machines could buy media in real time at enormous scale. It also squeezed publisher yields, added middlemen who never left, and gave buyers a supply chain they couldn't see into. Agentic buying is a chance to scale direct and programmatic transactions, and this panel asked what that will take.

"Programmatic was a huge innovation," Rio said, crediting it with faster transactions and automated planning. But "things haven't gotten better for publishers," because incentives were never aligned. Bosko named what was lost along the way: the best campaigns combine what a brand knows about its audience with what a media company knows about its context, and buying from a "soup of impressions" left no room for that conversation.

The panel agreed that agentic stands to restore relationships. "Agentic is about these buyer and seller conversations," Anthony said. Agent-to-agent communication can route around the middlemen and make deals more transparent. Agents aren't limited by human memory when they assemble an RFP response, and they can do it more efficiently than the manual process publishers have traditionally relied on. Humans still have the final say.

Behavior changes slowly, there's no wholesale disruption yet, and "there is AI in every pitch," as Rio expressed. Bosko's next step for the industry was interoperability, so agents can talk to each other regardless of which protocol is used.

Most contextual tools predate the AI revolution, yet many publishers have yet to rethink their approach in light of what AI now makes possible. These legacy solutions sort pages into broad categories, missing the meaning and nuance that make content valuable to buyers and sellers.

Optable’s Contextual Intelligence uses a large language model (LLM) to classify content, moving beyond those limitations to give publishers a richer understanding of what they publish—and more ways to turn that understanding into value.

With AI redefining what contextual technology can deliver, now is the time for publishers to reevaluate their approach.

How Contextual Has Traditionally Worked

Most legacy contextual systems in the market run on machine-learning models built years ago. They match patterns: which words show up on a page, how often, and next to which other words. Then, they assign the page a category.

That approach worked for a time, but it has limits:

• It reads words, not meaning. Pattern matching struggles with ambiguity, sarcasm, nuance, and passing mentions. To a model that only counts words, a headline like “Apple shares crumble” can look like a dessert recipe. For brand safety, that often means broad keyword blocklists that keep ads off perfectly safe pages.
• It often sees only part of the page. Many systems classify a page from its URL, headline, or a snippet. They don’t evaluate the full article and miss out on information like tone, timing, and context.
• It's built for one taxonomy. Traditional models are trained on a fixed set of categories. Adding a new or custom taxonomy means new data and another round of training, so many systems classify once and translate the result. Precision declines along the way.
• It stops at the page. Categories sit in one system and audience data in another. Buyers can target a page or a person, but rarely both at the same time.

That was fine while third-party identity carried most of the load. Now, contextual has to do more of the work, and the gaps are hard to ignore.

Optable Goes Beyond Keywords

Optable's Contextual Intelligence doesn’t just count words or filter content into surface-level categories. We built it on large language models, the same kind of AI behind today's leading assistants, to read the page for meaning. What that means:

• It understands context. The model picks up nuance, tone, and complex topics, so it can tell Apple the company from apple the fruit. When a single article spans more than one topic, like personal finance and automotive, it assigns each relevant category instead of forcing a single best match.
• It covers more than the topic. Every page is directly classified for IAB content categories (3.1 or 2.2), keywords, brand safety, sentiment, and emotion. Advertisers can trust the signal and rely less on broad blocklists that keep them off safe, high-quality content.
• It keeps up with new content. The model understands language broadly instead of relying on a fixed set of training examples, so new topics and terms don't require a rebuild. Pages are also reclassified on a regular schedule to keep results current and accurate.
• It turns context into audiences. Every classified page view is tied to the reader behind it. That means publishers can build first-party audiences like “readers who viewed personal finance content three times in the past 30 days.”

That’s where Contextual Intelligence goes beyond traditional tools. The same classifications work on the page in real time. Publishers can monetize one-time readers from their first visit, and across visits, so they can build audiences of readers who keep coming back to a topic. Those audiences can be combined with other first-party data and sent to Google Ad Manager, SSPs, DSPs, and other destinations.

Google Now Sends Categories by Default. What Sets Your Inventory Apart?

As of October 2026, Google Ad Manager automatically adds IAB content categories to bid requests by default. That’s good for the industry, but it means everyone is now speaking the same taxonomy. When every travel page sends the same IAB travel category, that category alone won’t set your inventory apart.

Google’s IAB categories only reach Google’s demand. Many publishers sell through Prebid and other partners outside their ad server, where those categories never appear. Optable’s Contextual Intelligence classifies every page with the same standard IAB categories, and Optable brings those classifications to Prebid so publishers can send one consistent signal to every demand partner.

In addition to IAB categories, Optable classifies every page for keywords, brand safety, sentiment, and emotion, and that depth matters more as buying becomes more agentic. Buyer agents are starting to read briefs, evaluate inventory, and recommend packages, but they work only from the signals they’re given. With richer context, an agent can tell a travel article about an exciting new ride from one about how people can’t afford to travel this year: same IAB travel category, opposite moments.

How To Get More From Contextual Today

Basic categories now come with every bid request, and agents are starting to do more of the buying. The advantage goes to those who act on richer signals first. Here’s where to start.

For Publishers

1. Audit your classifications. Check how 10 to 20 nuanced pages are categorized, and verify whether your provider reads the full article or just the headline. If pages are mislabeled or broad, test an LLM-based classifier on a few sites.
2. Connect content to audiences. Turn your readers’ favorite topics into segments for direct deals and PMPs. For one-time readers, pass page categories to your ad server as key-values.
3. Get agent ready. Run a recent RFP through a sell-side agent, like Optable’s Audience Agent or Sales Agent. If the agent falls back on broad sections, add depth to your signals with keywords and deeper categories.

For Brands and Agencies

1. Rethink your blocklists. Review the topics and keywords you block by default, like news or finance. With publishers that classify for meaning and brand safety, replace blanket blocks with page-level exclusions and win back safe reach.
2. Test the classification. Ask publishers for sample classifications of a few nuanced pages before you buy. If an article about Apple’s earnings comes back as food, that signal can’t be trusted.
3. Brief beyond the category. Add keywords, tone, and brand safety requirements to your briefs, not just IAB categories. Buyer agents can only match what you ask for.

The Next Evolution of Contextual with Optable

Optable's Contextual Intelligence brings large language models to contextual classification and connects the results directly to first-party audiences. Here's how it stacks up:

Publishers already running the Optable Web SDK can get started with no new integration. Those with an existing classification partner can connect it or import their classifications.

Contextual Intelligence reads the full page for meaning and turns what it finds into audiences publishers own. That gives buyers and their agents richer signals to work with, and it gives publishers inventory worth paying for, all on one platform with no new integration.

Want to see it on your own content? Schedule a meeting with the team.

Optable maintains that publishers' identity, audience, and pricing data serve as context for AI. That claim gets tested every time a buyer tries to connect its own data to a non-standardized range of publisher signals. This session put a builder of agents and a buyer of media on stage to share their perspectives.

Matt described the problem agents solve best: no person can read every signal across every surface, but an agent that reads the material, KPIs, analytics, and trends can surface insights and show lift immediately. He called AI tools "an upgrade from fixed modeling processes," and noted that the industry's focus has moved beyond time savings to better outcomes and more frequent measurement, as well.

Shukmei brought the buyer's view. Automation reduces manual effort, works around third-party dependencies, and produces measurable insights faster. "What everyone wants is a clean line and an understanding of identity," she said, and she argued for owning your data and your identity graph.

Both speakers drew a firm line on how to get there. "All of these decisions start with data," Matt said. Shukmei warned that "AI can't automatically fill in the gaps for you," and urged teams to verify outputs and add signals.

The takeaway for data, measurement, and planning teams: get the data foundation in order, then adopt agents strategically and measure often.

Recorded live at Optable Connect NYC 2026

The largest players in advertising keep consolidating toward end-to-end, owned infrastructure. Recent moves in the identity space have added pressure to go all-in on a single platform. The alternative is building every layer in-house, which costs time and money most agencies don't have.

Goodway Group chose a third path with Optable; Paul Frampton-Calero, Goodway Group's CEO, sat with Vlad Stesin, Optable's CEO, to discuss why they made this move and how it's performed.

Paul wanted an intelligence layer that works horizontally across the ecosystem: build once, then push to every partner and project. Instead of adding another workaround or building from scratch, Goodway built Passport One, its proprietary data and intelligence layer, on Optable's identity infrastructure. "You don't need to own everything yourself," he said.

Using Optable Audience Agent, Goodway goes from a brief to audience segments in around 30 minutes, work that could take days or weeks by hand. Some of those agent-built audiences delivered up to 40% better ROAS, according to an onstage comment from Paul. Planners can now spend their time on more strategy-driven tasks. Goodway has cut its SSP partners while adding direct publisher connections.

Paul also had advice for the sell side: Identify and prioritize the agency partners with an appetite for agentic advertising. For agency leaders deciding what to build, buy, or partner on, this conversation is a working example of owning the layer closest to clients and orchestrating the rest.

At Optable Connect NYC on October 8, we announced the Optable Agentic Marketplace, where buyers and sellers can transact directly via their agents. Then, we revealed that Optable would charge a 0% fee on direct deals between buyers and sellers.

That’s because the current AdTech ecosystem is a complex, confusing, and costly maze. Agentic buying could reduce the stack between buyer and publisher, or it could add even more tolls. 

We need to get agentic advertising right. But what does that mean? What does that look like? How about something like this. How about a seller, a buyer, and a direct connection between them? That should be our standard for whatever we build in this new ecosystem. — Andy Sharkey, VP of Product

Agents Shouldn’t Maintain the AdTech Tax

Of every $100 an advertiser spends in open programmatic, just $51 reaches the publisher. The rest is split among intermediaries, and about $15 can’t be traced to anyone.

That’s the nature of the AdTech tax — a stack of technology providers that siphon off the cash flow meant for the people selling their inventory.

Agentic buying could course-correct the pattern. A buyer agent can send a brief straight to a publisher’s seller agent, which proposes matching products at the publisher’s price. Once the publisher approves, the deal gets booked within the publisher’s ad server. An agentic transaction reimagines the buy, with direct deals at programmatic speed and scale.

The risk is that every intermediary bolts on its own agent and keeps its cut, so the same tolls get collected at machine speed. A take rate makes that worse. When a marketplace takes its cut from what the publisher is paid, the deal value shrinks.

We chose not to build that into the Agentic Marketplace, because every dollar a marketplace keeps is money that never reaches the publisher. In short: direct contracts between buyers and sellers incur no additional charge; buyers who want advanced features like clearing pay extra.

No Take Rate for Publishers; No Meter on Briefs or Buys

Agentic connectivity is free for Optable Audience customers. No activation fee, and no charge per task, brief answered, or buy booked. The Agentic Marketplace doesn’t need a toll, because the platform is what publishers pay for.

That leaves buyers and sellers with two ways to transact:

• Directly, between platform customers. Buyers and publishers contract with each other, the deal gets booked in the publisher’s own ad server, and each side invoices the other. Optable never touches the money.
• Through the Marketplace’s advanced features. Buyers who want additional features — such as transaction clearing, consolidated billing (one invoice across every publisher), buyer audience onboarding and targeting, and CAPI integrations — pay an access fee for those services. Optable handles the invoicing, and sellers still pay 0%.

Discovery is free for both sides. Optable’s revenue comes from software worth paying for, not a cut of what publishers earn.

Software should be paid for as software. — Vlad Stesin, CEO, Optable

What Buyers and Sellers Get Today

Onstage in New York, Andy Sharkey, VP of Product, and Chris Couture, Senior Product Manager for Agentic Advertising, walked a sample deal, end to end, through the Agentic Marketplace.

Representing the sell-side, Andy gave a brief tour of the Optable platform where sellers can build ad products (their inventory packaged against a price point). By using Optable as their storefront, sellers can make their inventory discoverable to buyers and their agents. So when a buyer submits a brief through the marketplace, they are presented with matching ad products from each consenting publisher.

Chris showed the Optable platform from the buyer’s perspective. A buyer can connect its buyer agent (including Claude, ChatGPT, or custom-built agents) to submit briefs to the Agentic Marketplace on its behalf. Then, the agent can identify media sellers that have opted in to being discovered by the buyer’s brand.

The buyer agent can then support campaign planning, with reasoning to back up its recommendations, and create a media buy for the planner to approve before sending it to the seller.

From there, the seller can approve the buy, send it to their ad server, and give the final say before it goes live. The Agentic Marketplace stands to shorten the timeline from brief to buy, reduce the number of intermediaries involved, and shrink the AdTech tax while restoring the direct buyer-seller relationships the industry was built on.

For sellers:

• Define products, pricing, and rules once, and buyer agents can discover them.
• Approve every buy and every creative before it runs. Approved buys book straight into a publisher’s ad server.
• Tap into new sources of demand, and see what buyer agents ask for, in their own words.

For buyers:

• Reach every publisher running a Sales Agent through one MCP connection, briefing from Claude, ChatGPT, or your own agent.
• Track every buy with delivery and performance across publishers in one view.
• Use seller audience data and insights or bring your own audiences to premium publisher inventory.

Coming next: A custom ad server connector and proposal management for sellers, plus buyer data rails and more for buyers.

Are You Ready For Agentic Advertising?

Of the attendees we surveyed at Optable Connect, more than half expect to run an agentic deal by the end of 2027, and another 25% have already run one. We built the Agentic Marketplace as a gateway to agentic buying and human-approved selling.

For publishers, the first step is small: pick one high-value package, price it, and put it in front of buyer agents. For agencies and brands, bring a live brief and connect the agent you’re already using.

The Optable Agentic Marketplace is available to pilot today. Talk with the Optable team to get started.

The lobby hummed with warm, friendly greetings and enthusiastic introductions as nearly 200 leaders from all sides of the advertising space converged at the People Inc. Auditorium in Lower Manhattan for Optable Connect NYC. Reunions and new connections alike shared in the early morning excitement over coffee, pastries, and the golden sunlight of a perfect October day.

Screens depicting a packed 8+ hour agenda complemented and endorsed the playful animations proclaiming that “Agentic advertising runs on Optable.” Throughout the hallway, brightly colored blocks seemed to bounce off the floor, drawing the eye and confirming that the forward-thinking, agentic-inclined attendees were in the right place.

As the clock inched towards showtime, the quiet auditorium saw a slow trickle of attendees that quickened into a flood, filling the space with soft murmurings of interest and anticipation. Then, with nearly 200 seats filled, the program could finally begin.

Patrick Viau, Optable’s CRO and emcee extraordinaire, opened the event with a congenial attitude, welcoming folks from the buy and sell sides, press organizations, analyst desks, and beyond. He quickly addressed the agentic elephant in the room: the buzzword that is “agentic” itself. But with the kind of confidence that easily erases doubt, he declared, “Agentic advertising is here.” The proof: people are already transacting agentically, and that pattern isn’t likely to slow down any time soon.

The day ahead would further support this claim, as agency representatives, publishers, platforms, journalists, analysts, and researchers covered how agentic deals restore the buyer-seller conversation, why data foundations come first, where humans stay in the loop, and how agentic grows alongside programmatic.

Optable’s CEO, Vlad Stesin, took the stage next, with a note of appreciation for Patrick’s energy and the transformative effect he’s had on the company since joining, as well as a slight condemnation of Patrick’s alleged propensity for hugging. Vlad then built on the foundation Patrick laid, expressing that “advertising is about to be run by software that talks to other software, on top of software it trusts.”

That’s the future of advertising Optable is building towards.

The Case for Agentic Advertising

Earlier this year, the speakers at Optable Connect Toronto agreed that agentic advertising had arrived. Those of us in New York last week tackled some important follow-ups: what does that look like in practice, and what does the market need in place, on both sides, to make it work?

Vlad's answer started with what programmatic left behind. Scale came at the cost of context, and with it the direct conversation between advertiser and publisher. “Agentic advertising puts context back into the transaction,” he said, and with it, direct deals “stop being a privilege of the biggest players.”

He was just as clear that technology alone won't carry the shift: “Disruption isn't about the availability of new technology. It's a change in behavior.” Optable's role, in his framing, is to bring order to signal chaos: “We're building the foundation for a new era of transactions.”

Next to the stage: Co-Founder and CPO Bosko Milekic, VP of Product Andy Sharkey, and Senior Product Manager Chris Couture. Bosko opened with the question Optable has asked all year, “Are you agent-ready?”

Andy described agentic advertising as direct IOs with the success of programmatic, where Optable is the storefront for sellers to create their ad products. He and Chris walked through an end-to-end media buy in which agents share data without exposing it, plan the campaign, and prepare the order. Every step up to the buy is automated, and nothing goes live until a person approves it.

The announcement that drew the biggest reaction: the launch of the fee-free Agentic Marketplace for buyers and sellers, with access fees for buyers who want advanced features.

From there, the agenda widened from Optable's roadmap to the market as a whole. A few threads ran throughout the day’s sessions, across media buyers, sellers, analysts, and industry experts. Each one is worth considering for your 2027 planning.

Agentic Deals Restore the Advertiser-Publisher Conversation

“Programmatic was a huge innovation,” Credera's Rio Longacre said during a session on what agentic can learn from programmatic. Programmatic scaled the transaction and automated the planning. Bosko named what the innovation cost: the best campaigns combine brand and media company knowledge, and that conversation had no place in an automated auction. Buying from a “soup of impressions” has left little transparency, and advertising has been optimized for volume over impact.

Agentic stands to bring the conversation back. “Agentic is about these buyer and seller conversations,” said AdExchanger's Anthony Vargas. Mediavine CRO Amanda Martin put it in human terms: “This industry runs on relationships,” and when buyer and seller agents talk, buyers and sellers are talking. 

EMARKETER's Yory Wurmser added a market view: AI will first benefit the walled gardens, he said, but agentic, as an extension of direct demand, can level the field for smaller players, as well.

Restoring the conversation also changes the economics. Rio said that “things haven't gotten better for publishers” since programmatic because incentives aren't aligned, and RTB supply keeps rising while publisher traffic falls. Amanda's ask of the buy side was direct: “We need the buy-side to push the middlemen out.” Goodway Group is already moving that way, with fewer SSPs and more direct publisher connections since building Passport One on Optable.

Data Foundations Decide Who’s Ready for Agents

If one piece of advice came out of nearly every session, it was this: agents are only as useful as the data underneath them. “All of these decisions start with data,” Newton Research's Matt Emans said. He warned that throwing an agent at any task “isn't always the best course of action” until you understand the data and methodology behind it.

Involved Media's Shukmei Wong agreed, expressing that users shouldn’t blindly trust AI outputs, especially if they haven’t vetted their data first. What everyone wants, she said, is “a clean line and an understanding of identity.”

The room's own answers backed them up. In the live survey, more than a quarter of participants said their data and systems aren't ready for agentic buying and selling. Google Cloud's Bora Beran made the same point from the infrastructure side: “You have to provide clean data and enough context to the model,” and a good foundation makes everything after it easier to scale. Yory's checklist for agentic readiness included getting data organized in a way agents can use.

The day's final panel showed what that looks like for publishers. Panelists noted that the way inventory is described today, with simple labels like “sports” or “lifestyle,” misses the richer, more nuanced signal buyers want.

Amanda's advice: “Invest in the data foundations,” starting with a deep dive into your own signal treasure trove. Hearst's Dave Lundell described that foundation as ongoing work, built for what you're doing today and what comes next. It's also where Optable started the day: Vlad described Optable Audience Agent as a way of “turning passive data into an active asset.”

Agents Do the Legwork; Humans Approve the Deal

Goodway Group CEO Paul Frampton-Calero shared the impact of strategically adopting agentic tools. Using Optable Audience Agent, he reported that his team goes from a brief to audience segments in under an hour, work that could take days or weeks by hand, and some of those agent-built audiences delivered better ROAS than the alternatives. The time saved goes back into strategy. Matt Emans described the same shift across the industry: 12 to 18 months ago, the focus was on time savings, and now it's on better outcomes and measuring more often in addition to efficiency.

The need is just as clear on the sell side. “A week between brief and buy is optimistic,” Amanda said of today's direct-deal process. Unity's James Prudhomme noted that a model “can consider a much, much wider set of audiences” than a manual workflow.

None of that takes people out of the deal. Two-thirds of survey respondents said they want agents to have more freedom to negotiate and prepare orders, with people finalizing and approving them, not full autonomy. The workflow Andy and Chris demoed still ends with a person approving the order.

Heather Carver, of tvScientific by Pinterest, urged attendees to treat AI as “a copilot, not on autopilot,” and Shukmei urged teams not to assume an LLM's first answer is correct. On the sell side, speakers on the signal panel argued that human-in-the-loop makes sales teams more valuable, since relationships still close the deal.

Early But Not Premature: Agentic Grows Alongside Programmatic

Of the 105 attendees who participated in the live survey, 76% lean positive on agentic advertising, and 94% have already run an agentic direct order or plan to by the end of 2027. For most, though, that first order is still ahead: 51% expect it in 2027.

The most-cited hurdle was people, not tools: 37% named internal organizational readiness as the factor holding them back from agentic adoption. “It takes a long time for people to change the way they behave,” Yory said, and Rio agreed: there's no wholesale disruption yet, but “there is AI in every pitch.”

What's taking shape is a layer that works alongside programmatic. Yory pointed out that agents can improve programmatic, too, with better signals, faster decisioning, and more transparency. IAB Tech Lab CEO Anthony Katsur described AAMP as an extension of existing standards, and said agentic standards are being built to support both direct and programmatic buys.

Paramount's Derek Nicol reminded the room, “We're an industry built on great standards.” With AAMP and AdCP increasingly overlapping, Yory and Bosko both argued that interoperability matters more than which protocol wins.

Early doesn't mean there's time to wait. During the survey, a buy-side attendee warned that buyers aren't holding off while publishers wait for the money to show up first. In their view, publishers who build now will be part of the agentic equation, and those who wait risk being left out. Paul's advice to publishers was practical: identify and prioritize the agency partners with an appetite for agentic.

Session-by-Session Breakdown

Here's a closer look at what each session brought to the stage, picking up threads from the themes described above.

Own Less, Orchestrate More: Partnering Without Owning the Stack

Vlad Stesin, Optable, with Paul Frampton-Calero, Goodway Group

Paul explained why Goodway wanted an intelligence layer that works horizontally across the ecosystem, and why it built Passport One on Optable instead of trying to own every piece. 

“You don't need to own everything yourself.” — Paul Frampton-Calero

With Optable Audience Agent, his team now goes from brief to audience segments in 30 minutes instead of days or weeks, and some of those audiences reportedly delivered 40% better ROAS.

Ground Truth for Agents

Vlad Stesin, Optable, with Matt Emans, Newton Research, and Shukmei Wong, Involved Media

Matt and Shukmei agreed that agentic decisions start and end with data, and that agents deliver the best results when teams understand the data and methodology behind them.

Matt described agent pipelines that read signals across surfaces to surface insights a person would miss, while Shukmei cautioned that “AI can't automatically fill in the gaps for you.”

From Plan to Reconciliation: Inside IAB Tech Lab's AAMP

Anthony Katsur, IAB Tech Lab, and Derek Nicol, Paramount

Katsur explained how IAB Tech Lab's Agentic Ad Management Protocol (AAMP) builds on existing standards and taxonomies to support both direct and programmatic buys, and how IAB Tech Lab's “early, but not premature” agent registry will add trust and transparency. Nicol called standards “the building blocks of success” and urged attendees to ask their vendors how they’re incorporating AI into their product roadmaps.

From Programmatic to Agentic: How AI Could Re-wire Media Buying

Yory Wurmser, EMARKETER

Yory argued that AI will first benefit the walled gardens, but agentic, as an extension of direct demand, can level the field for smaller players and make programmatic more efficient.

His advice: be specific about what you want from agentic, set clear governance rules, get your data ready, and experiment.

What Agentic Buying Has to Learn From Programmatic

Anthony Vargas, AdExchanger (moderator), with Rio Longacre, Credera, and Bosko Milekic, Optable

Rio acknowledged how innovative programmatic has been but said incentives aren't aligned, leaving publishers stranded. Bosko described what agentic has the potential to restore: a conversation between what a brand knows and what a media company knows. He called for interoperability so agents can talk regardless of which protocol wins.

Are You Agent-Ready? A Live Audience Survey

Andrew Dumas and Andy Sharkey, Optable

Andrew and Andy polled the room live on agent readiness and agentic advertising. 105 attendees answered live, most representing publishers and sell-side tech. Of the respondents, more than three-fourths lean positive on agentic advertising. In fact, 94% have already run an agentic direct order or plan to by the end of 2027.

Between questions, one buy-side attendee recommended publishers keep up with the agentic shift. Another attendee increased the pressure, highlighting their own success in running agentic deals.

How Unity Made Its Gaming Audience Addressable and Omnichannel, With Optable and Google Cloud

James Prudhomme, Unity, Bora Beran, Google Cloud, and Marc-Olivier Thibault, Optable

James made the case that stable addressability future-proofs Unity's business, and described using Audience Agent to build segments from a far wider set of audiences than human intuition could consider. Google Cloud's Bora Beran described AI as a companion that works only as well as the data and context it's given.

Making the Most Valuable Signals Easier to Buy

Crystal Jacques, Optable, with Heather Carver, tvScientific by Pinterest, Dave Lundell, Hearst, and Amanda Martin, Mediavine

Each panelist described their most valuable signal — Pinterest's commercial intent across visual search and shopping, the trust of Hearst's readers and viewers, and Mediavine's contextual data — and agreed that flattening it for programmatic loses the context buyers want. Their shared advice: invest in your data foundation, and describe your inventory so agents can read it.

Closing Remarks

Andrew Dumas, Chief Business Officer, Optable

Andrew pulled the day together, describing Optable as “a company of builders” that builds “brick by brick,” highlighting the “freegentic” model announced for Optable’s Agentic Marketplace. He thanked speakers for contributing, encouraged guests to connect with the team, then sent everyone out to the terrace. 

Next Steps for Agentic Advertising

Whether you buy media or sell it, most of the work happens before an agent ever touches a campaign. Based on what we heard, here's where to start this quarter:

• Audit your data foundation. Find out what valuable signals you have, evaluate how they’re structured, and adjust so an agent could read them. 
• Pick one workflow to automate, with a human approval step. Decide whether to build, buy, or partner on it this quarter, and name who owns it.
• Get fluent in the standards. Know where AAMP, AdCP, and MCP each fit in your workflow before a partner or buyer asks.
• Ask your partners where they stand. Find the agencies with an appetite for agentic, and ask your vendors how AI fits in their roadmaps.

Optable is Building the Infrastructure for What Comes Next

Vlad opened the day with a picture of advertising “run by software that talks to other software, on top of software it trusts.” Over the sessions that followed, the room heard what that trust rests on: data that agents can read and transact against, with a person approving every deal.

That's the foundation Optable is building.

Optable provides the identity infrastructure, data collaboration, and agent-to-agent tooling, including Optable Audience Agent and Sales Agent, that let publishers and agencies act on these ideas.

Earlier this year, we shipped our agentic signals layer to put them to work. Now, we’re inviting publishers, platforms, agencies, and advertisers to participate in our Agentic Marketplace, where buyer agents and seller agents can transact directly — faster deals, better-fit audiences, and buyers and sellers back in conversation.

Wherever your team is on agentic, we're ready to build with you.

Thank you to every speaker, partner, and attendee who closed out Advertising Week New York with us and made the trip worthwhile.