From prompts to real orders: helping publishers get agent-ready
BY HAZEL BROADLEY, BEELER.TECH
Agentic AI is one of the most talked-about topics in advertising right now, and one of the least understood. An LLM can happily draft an RFP response in seconds, but turning buyer demand into repeatable, revenue-generating campaign orders is a much bigger job. The gap between those two things is where most publishers currently sit.
To unpack it, Rob Beeler sat down with Andy Sharkey, VP of Product at Optable, an agentic audience and data collaboration platform. Their conversation covered where agentic advertising is genuinely useful today, the data foundations a publisher needs before it can trust an agent, the organizational shifts required to adopt it, and what a real buyer-to-seller workflow actually looks like.
Here are the key takeaways for publishers from this exclusive community conversation.
Recognize the hype, the reality, and everything in between
Andy splits agentic advertising into three buckets, and only two of them should be on a publisher’s radar.
The first is custom algorithms, including the IAB Tech Lab’s Agentic Real-Time Framework. That containerized execution model runs alongside an SSP and exists for the benefit of agencies and buyers, so publishers have very little to lean into here. If you’ve been fielding nervous questions about it, the honest answer is that it isn’t built for you.
The second is campaign-planning agents, and these are already earning their keep. A seller can drop an emailed RFI or RFP straight into Claude, ChatGPT, or Copilot and get back a proposed response, with a matching audience, suggested inventory, and pricing to help you reply quickly. Publishers are adopting this today.
The third is campaign-selling agents, where the buzz is loudest and the reality is thinnest. By connecting buy-side and sell-side agents, in time these could become an alternative, or even a competitor, to RTB. Plenty of startups have sprung up to test the idea, and plenty of press releases have followed. But those releases still celebrate a single campaign, precisely because very few orders and very little real budget are moving through those pipes yet. The interesting questions are why, and what needs to change.
Get your data house in order
An agent is only ever as good as the data underneath it. Andy uses a neat analogy: ask three different LLMs to summarize a long YouTube video and Gemini wins, simply because Google holds the underlying data the others can’t see. The lesson for publishers is that being agent-ready starts with three core signals.
Identity comes first, and it’s far more than whether a user carries a cookie. The goal is a cross-platform view: cookies where you can get them, mobile advertising IDs to extend into apps, IP signals to help connect CTV, all stitched into a graph that keeps data interoperable as it moves between your audience platform, DMP, SSP, and ad server. Clear provenance is what lets a seller answer an agency’s questions with confidence, rather than wondering whether the model just invented something.
Contextual intelligence needs to go beyond keywords and the IAB content taxonomy. Buyers now want sentiment, emotion, people, and places, plus brand-safety nuance. Older machine learning got the industry this far, but an LLM is the strongest tool yet for understanding what a page is really about. If you already work with a contextual vendor, it’s worth checking what actually powers their classifications.
Audience rounds out the set. First-party behavior should anchor it, but the smart move is preloading a few thousand off-the-shelf third-party segments so you’re ready to activate the moment an RFI lands. That beats the old workflow of running out to a data marketplace, buying something, syncing it, and activating it every single time a gap appears. Load up once, then treat it as an all-you-can-need resource.
Remember, ‘use AI’ is not a strategy
Here’s where a lot of publishers get stuck. The first conversation with Andy is about technology. The second is usually, “we’ve built the tech, so why aren’t we getting traction?”
The answer is rarely more data. It’s people.
Moving a business into the AI world takes alignment from the executive team all the way down to operations, with product, engineering, and sales pulling in the same direction. That means new policies, new workflows, and sometimes a rethink of how the org is structured. A blanket “everyone should use AI” directive from the top is one of the least useful things a leader can say, because it skips the harder questions of who you are, who you serve, and what you’re actually selling.
The relationship with the sales team is the one that comes up most. Agents should act as a catalyst for relationship-led selling, helping good salespeople sell more and faster. Protected rate cards are a good example of where control has to stay with the seller. Publishers won’t, and shouldn’t, expose every price or transact blindly, and any credible vendor needs to respect that from day one.
Explore agents in action: from Clorox brief to live campaign
To make it concrete, Andy walked through Optable’s workflow using Claude as the buyer-facing interface, though it works equally through ChatGPT or Copilot.
Playing the buyer, he uploaded a mock Clorox brief and asked for audiences that would hit the campaign’s KPIs. Claude interpreted the brief, passed it to Optable’s technology, and returned a menu of options from multiple sellers: behavioral audiences, display or video, web or app, CPMs, the exchange or ad server each was available on, device targeting, even a deal ID ready to load into a DSP. In a live scenario, the next click turns that selection into a real campaign.
The seller side looks completely different, and deliberately so. Optable’s sales-agent interface lists multiple ad servers, including GAM, FreeWheel, SpringServe, and Publica, because publishers rarely work with just one. As more SSPs and ad servers build their own agents, no publisher wants to juggle 50 of them. An orchestration layer solves that by talking to all your ad servers and SSPs, and to their agents where they exist, so you get one interface instead of a mess of them.
From there, you define what’s for sale. Maybe that’s a homepage takeover at an $8 CPM as a standard line item. When a buyer’s brief matches your signals and inventory, that product surfaces in their results. Crucially, you don’t have to give everything away on day one. Carve out ten ad units to experiment with, keep the rest old-school, and learn as you go. And because Optable isn’t an ad server or an SSP, it stays agnostic about where the order runs.
That decision is yours.
There’s a hidden upside here, too. Even when you don’t win a campaign, the brief tells you something you didn’t know yesterday, such as a buyer hunting for Knicks fans over 40 in the tri-state area. That’s marketplace intelligence programmatic that’s never normally handed to you, and it can shape your content and audience strategy so you win the next one.
What publishers should do now
None of this is a light lift, but the hard work pays back. Andy’s advice comes down to three moves.
First, get your data in order, starting with identity, then contextual classification, then supplemental audiences ready for activation. Second, decide how a sales agent will represent your inventory to buyers, whether you build it yourself, lean into the open-source Prebid project Optable is helping develop, or partner with a vendor. Third, create alignment across the organization, from the leadership team to operations, on what being an AI-driven business really means for your sales team.
The prize is worth the effort. A more direct order model could sidestep programmatic’s longest-running problems, trim the ad tech tax, and share the economics more fairly between the buyer, the brand, and you as the content creator. Instead of being told what an audience you can’t see is worth, you get a platform to understand your own audience and take it to market on your terms. The aim isn’t more proposals sitting in a drawer, it’s real campaigns generating real revenue, hopefully starting in the second half of this year.
Ready to turn agentic planning into real orders?
If your team is ready to strengthen its data foundations and turn more RFPs into workable audience plans, Optable can help you find a practical route forward. Its platform connects the workflows behind audience planning and activation while keeping you in control of your data and your commercial decisions. Contact Optable to talk through your goals and see how agent-ready selling could work for your business.
👉Want to stay in the loop with what’s happening in the Beeler.Tech community? Subscribe to our newsletter. If you’re interested in attending or sponsoring a future event, you can explore our upcoming events here.
This is content created in paid partnership with Optable. We only feature partners who we believe bring real value to the publisher community.