Episode 187: Jack Raines Trolls LinkedIn and Ends Up with a Book Deal

Full analysis

Google wants publishers and agencies to negotiate and run campaigns directly through AI tooling wired into Google Ad Manager, its dominant ad server. Ari Paparo and Paul Knegtgen flagged a Google-sponsored DigiDay piece as the evidence: REST APIs plus an MCP server (a plug that lets AI agents talk to GAM directly) that could let buyers and sellers transact without an independent DSP or SSP in the middle. That's the story worth an operator's attention. The rest of the news block is real but lower-stakes.

What's actually being decided: not one company's move, but whether the plumbing of open-web programmatic gets routed around by AI-mediated direct deals. Type 1, hard to reverse, if it takes. Forcing function is Google's own API/MCP rollout plus a still-pending antitrust remedy.

The council

The Market Analyst. Paparo's read on the antitrust angle is the interesting part. The government wants Google to divest AdX, its ad exchange. Paparo's point: if buying moves to AI-powered direct deals inside GAM, AdX matters less, so Google could hand over the exchange and keep the ad server that actually locks in the market. The remedy gets aimed at yesterday's chokepoint. Plain version: the regulator is fighting over a toll booth on a road people are about to stop driving. For independent DSPs and SSPs, that's the threat. If the transaction happens agent-to-agent through GAM, the layer that skims the middle gets thinner. Watch Trade Desk framing here, it's the named DSP that gets bypassed in this model.

The Skeptic. Slow down. A Google-sponsored article promising REST APIs and an MCP server is a roadmap, not a product. "Negotiate campaigns through AI tooling" is a demo sentence. The gap between an agent that fills a form and an agent that actually clears a media plan at scale, with brand safety, frequency, and reconciliation, is enormous. Paparo's Higgsfield claim, half of all Reels made with it, he hedged with "I would guess," and it has no basis. Same energy across the block: big directional claims, light evidence. "Ad networks are back" is a vibe, not a P&L. Don't rebuild your stack around a sponsored blog post.

The Operator. Say Google ships the GAM APIs tomorrow. Who at a mid-size publisher actually wires an AI agent into their ad server and lets it negotiate? Nobody, for a while. The first thing that breaks is trust: an agent that misprices a direct deal costs real money, and no ad ops lead hands that over unsupervised in year one. But the second-order effect at 90 days is the quiet one. If Google's tooling makes direct deals cheap to execute, your SSP integration stops earning its rev share, and you start asking why you pay for a layer the ad server now does. That erosion is gradual, not a switch.

The Customer / End User (the agency buyer). The WPP story is the one that actually touches a buyer's day. Sony alleges WPP took roughly $350 million in Chinese media rebates and passed back about $110 million, keeping around $240 million through shell companies. Those are plaintiff allegations, unproven, in active litigation. But the reason it lands: rebate opacity is exactly why big advertisers are dragging spend toward retail media and direct platforms where the money trail is shorter. Walmart Connect growing 38% year over year isn't unrelated. Buyers want fewer hands between budget and inventory, and every rebate headline makes the agency layer look like one hand too many.

The CFO. LiveRamp is the clean lesson. Shareholders rejected a roughly $98 million executive pay package tied to the acquisition. Paparo thinks the deal closes anyway, but the acquirer now inherits retention risk with the parachutes voted down. For any operator eyeing an exit, that's the read: the 280G golden-parachute math (the tax rule that governs change-of-control payouts) is getting contested by shareholders, and a rejected package doesn't stop a deal, it just leaves the buyer holding the problem of keeping the people who built the thing. Price your integration risk accordingly.

The tensions

Is the Google story a market-structure shift or a slide? The Market Analyst treats agent-to-agent buying through GAM as the thing that makes the antitrust remedy toothless. The Skeptic says it's a sponsored roadmap and the hard part, agents that actually clear complex media at scale, is years out. Both can't be near-term right.

Do the news items connect or just co-occur? The Customer sees a through-line: rebate scandals, retail media growth, and AI-direct all point the same way, buyers stripping out middle layers. The Skeptic says that's a narrative stitched over four unrelated stories on a light news day.

Synthesis

This hinges on one belief: does AI-mediated buying actually collapse the DSP/SSP middle, or just add a new interface on top of the same plumbing? The council leans toward "real direction, wrong timeline." The pressure on the middle layer is genuine and it predates AI, rebate opacity, retail media, and direct deals were already squeezing it. Google wiring agents into GAM accelerates that pressure; it didn't start it. But the near-term product is thin, and no operator hands negotiation to an agent in year one.

What to de-risk: don't restructure around the GAM announcement, do audit which of your SSP/DSP integrations still earn their rev share on their own merits, because that's the layer that erodes whether or not Google's agents ever ship. And if you're selling, assume shareholders now contest the pay package.

Prediction: Google will ship the GAM REST APIs and MCP server to general availability during 2026, but through the 2027 upfront and budget-setting season, no top-10 advertiser or holding company will run material open-web media plans agent-to-agent through GAM without a DSP in the loop.

Confidence: Medium. The tooling is real and coming; adoption at scale is the slow part.

Why: Google's sponsored signaling tells you the plumbing is being built, and Google ships infrastructure it commits to. But an AI agent negotiating and executing a real media plan has to clear brand safety, frequency capping, and reconciliation that a form-filling demo skips entirely, and no agency hands unsupervised spend to a counterparty's agent in year one when a mispriced deal costs real money. The reason the bypass-the-DSP story overshoots is that buyers keep an independent layer precisely so the seller's ad server isn't grading its own homework, and the WPP rebate mess is a live reminder of what happens when buyers trust the middle too much. The opposite outcome, agents clearing scaled direct deals inside GAM by the 2027 upfront, would require agencies to abandon that check faster than any tooling has ever been adopted in this industry.

Revisit by 2027-10-01: We're right if, after the 2027 upfront, major buyers are still routing open-web programmatic through independent DSPs rather than negotiating agent-to-agent inside GAM. We're wrong if a top-10 advertiser or holdco publicly runs material open-web spend directly through GAM's AI tooling with no DSP in the loop.

One more, because it's the cheaper call: the antitrust remedy will keep aiming at AdX while the ad-server lock-in stays untouched, which is Paparo's point and the more durable structural fact in the whole episode.