The most useful counterweight to agentic hype this week came from an agency actually doing the work. Goodway Group’s Andrea Kwiatek described, in AdExchanger on 14 August, a two-track reality: the agency is feeding campaign briefs to buyer agents and evaluating seller-agent recommendations — genuinely testing the agentic stack — while simultaneously reducing its SSP partnerships through rigorous supply-path-optimisation validation.
Her framing of the second track is the sentence publishers should pin above the desk: “SPO is not about cutting out partners. It is more about understanding what those partners are able to support in terms of outcomes.”
And her warning about the first track completes the picture: agentic buying, she cautions, risks becoming “another media-buying black box” without agent-to-agent transparency.
Read separately, these are two industry commonplaces — agencies test AI, agencies trim SSPs. Read together, they describe the environment your supply is about to be sold into: the standard Kwiatek articulates — what can this partner evidence in terms of outcomes — is being applied right now by humans running SPO spreadsheets, and it is about to be applied continuously, at machine speed, by the buyer agents the same agency is training on its briefs.
01The dropping is automated now
The traditional SPO cycle had human friction built in. A path review happened quarterly or annually; a publisher or SSP that fell short got a warning, a meeting, a chance to present; relationships bought time. Buyer agents remove the friction. An agent evaluating seller-agent recommendations against a brief does not schedule a review cycle — it scores paths on every decision, and a path that cannot evidence its fees, hops and outcomes does not get argued out of the plan. It simply stops being selected.
This is why Kwiatek’s gentler formulation — understanding what partners can support, not cutting partners — is not actually gentler for the unprepared. The standard is evidentiary. Partners who can document what they support pass; partners who cannot are indistinguishable, to an algorithm, from partners who support nothing. The dropping becomes faster and less forgiving than the human kind, and there is no lunch at which to contest it.
The black-box warning cuts in publishers’ favour, though, if they move. Buyers burned by opaque algorithmic products — the theme running through this week’s AppLovin and Wall Street coverage — are precisely the buyers demanding agent-to-agent transparency. Supply that arrives documented — clean path, declared fees, verifiable outcomes — is the answer to the anxiety Kwiatek is voicing. The opening is real: transparent supply becomes more valuable as the buying layer gets more opaque, because it is the part of the chain a nervous buyer can still inspect.
02Your machine-readable representation is your sales deck
There is a second-order shift buried here. When a buyer agent evaluates seller-agent recommendations, what it evaluates is not your media kit or your sales narrative — it is whatever machine-readable representation of your inventory your SSPs surface to it. Sellers.json entries, ads.txt declarations, supply chain objects, whatever outcome and quality signals the SSP passes through: that is the version of you the algorithm shortlists or skips. Most publishers have never seen that representation, never audited it, and could not say today which of their SSPs presents them well or poorly to an agentic buyer. That is roughly equivalent to never having read your own sales deck.
The agentic future and the SPO present are the same discipline at different speeds, and Goodway Group is usefully honest about running both at once.
03Why this matters for publishers
| The evaluation standard is already set, and it is evidentiary | "What can this partner support in terms of outcomes" is the question your supply will be scored on — by SPO analysts today, by buyer agents next. Assertions do not score; documentation does. |
|---|---|
| Consolidation is happening on both sides simultaneously | The same agency testing seller agents is actively cutting SSPs. Fewer paths will carry more spend, and inclusion in the surviving paths is decided by exactly the evidence most publishers have never assembled. |
| Automated dropping has no appeals process | The human SPO cycle gave underperforming paths warnings and meetings. An agent just stops selecting you, and the first symptom is a revenue line drifting down without a phone call to explain it. |
| Opacity anxiety is your opening | Kwiatek's black-box warning means sophisticated buyers are actively looking for inspectable supply. Publishers who can evidence fees, hops and outcomes are selling exactly what the buy side says it fears losing. |
04What publishers should do
05The bottom line
The agentic future and the SPO present are the same discipline at different speeds, and Goodway Group is usefully honest about running both at once. The standard does not change when the buyer becomes software — it just gets enforced without meetings, without warnings, and without the relationship capital that used to buy underdocumented supply another quarter. Publishers who can evidence their path — fees, hops, outcomes, and a clean machine-readable representation — get shortlisted by whatever is doing the choosing. The ones who cannot get dropped without a meeting, and find out from the revenue report.