Part Two: Publishers Are Solving for a Specification No One Published.
The Connected Advantage: Build Together, or Drift Apart.
By Stuart Bailey | Part Two of Three
First published in B&T, 23 September 2026.
In Part One I wrote about advertisers automating a structure they never fixed. This piece turns to the other side of that same gap. Publishers are living the asymmetry too, just further from where the decisions get made, and with less say in how they're being judged.
No Shared Language, Anywhere
The five major holding groups have each built their own agentic layer over the past eighteen months. WPP Open. Publicis CoreAI. Omnicom Omni. Dentsu Connect. Havas Converged.ai. Five proprietary systems, five different sets of planning logic, five different sets of criteria for what makes a publisher's inventory worth recommending. None of them published. None of them talking to each other. None of them talking to you.
The independent market adds a different version of the same problem, not a smaller one. Australia's independent agencies don't share a common platform at all, each runs its own mix of planning tools, workflows and evaluation criteria, built around its own client base. There's no single system to learn, which sounds easier until you realise it means there's no shortcut either, no shared spec to reverse-engineer even once. Every agency relationship, holding group or independent, is its own integration problem.
Nobody outside those businesses, of either kind, was in the room when those decisions were made.
Where The Barrier Actually Sits
It would be easy to read the adoption gap as a willingness problem, publishers dragging their feet. The data says otherwise. Research from Digiday and Optable, surveying 180 publishers, advertisers and agencies in May 2026, found the two biggest barriers to publisher adoption of agentic AI are technical complexity (68%) and internal organisational readiness (51%), well ahead of unclear industry standards (41%) or regulatory uncertainty (37%).
Publishers aren't the reluctant party here. They're the party facing the harder integration problem, with the least visibility into what they're integrating with.
Getting Found Only Gets You In The Room
The same research showed only 5% of publishers are focused on building their own AI agents, while 49% are focused on improving first-party data and signals so external systems can find and read them.
Read on its own, that looks like the whole answer: get legible, job done. But, it's one of three things publishers need, not the whole answer.
The first is legibility: being readable to whichever system is doing the discovering, whether it sits inside a holding group or runs on an independent's own stack. The second is friction, or the lack of it. A system-to-system buying process has no patience for anything short of standardised terms, API-ready ad ops and fast turnaround. The third is fluency: understanding how your agency partners plan and what the advertisers measure success on. Sometimes that's different from what the brief says.
Fluency is the one that unlocks what publishers do best. AI can generate ideas at extraordinary speed. What it cannot reliably supply on its own is the contextual judgement that makes an idea distinctive, appropriate and commercially useful. Publishers are better placed than almost anyone to bring that context and creativity to the table. But a great idea aimed at the wrong objective still misses. Fluency is what stops that happening.
Being found gets you in the room. Being frictionless keeps you in the room. Fluency is what gets you chosen.
A Pattern Worth Naming
This asymmetry isn't a new problem. We saw the same pattern in the walled gardens era, when a handful of platforms built data and decisioning power no publisher could match, based on rules only the platform understood. I wrote about that precedent at length in the first piece of the Agentic AI Asymmetry Dilemma series.
This time is different, though, because of who's building it. Agencies sit between advertisers and publishers rather than owning the whole relationship the way a platform does, and their systems get better with every source they connect to, not every source they shut out. That gives agencies a real reason to open up rather than close off. Nobody's acted on it yet. But the reason is there.
Build All Three
The instinct is to fix one thing first, usually legibility, and worry about the rest later. That's the wrong order. All three need building at once, whether the agency in question is a holding group or an independent.
A publisher who's legible but not fluent is easy to find and easy to overlook. A publisher who's fluent but slow to close is right for the brief and too late for the buy. A publisher who's fast and legible but generic could win some transactions but lose the relationship.
So, you need to be found, frictionless, and fluent. Drop any one of the three and the other two stop working.
Score Yourself Honestly
So, score yourself honestly, not ambitiously.
Score yourself on all three right now, not just the one you've already made progress on. Legibility work doesn't excuse a gap in fluency, and a good relationship doesn't excuse a slow back office. The three don't trade off against each other, which means neither should your attention.
In the next piece we turn to agencies, and the mirror-image risk on their side: a system built well for one relationship that still adds up to disconnection across the whole ecosystem.
