Sunday night, Publicis announced a $2.2 billion all-cash deal to buy LiveRamp. The press release used the phrase “smarter agents” twice in the first paragraph.
Tuesday morning, Google said AI Mode crossed one billion monthly users and rolled out Ask Advisor across Ads, Analytics, and Merchant Center.
Friday, Criteo confirmed that more than 1,000 brands are now running ads through the OpenAI advertising pilot inside ChatGPT, with AI-attributed conversion rates running close to twice traditional channels.
One week. Three announcements. Same story.
B2B buying is about to go agent-to-agent. You’d better get ready.
Here’s the part nobody on a vendor stage is telling you: your buyer doesn’t have an agent. They might never get one. And every dollar your team spends preparing for that future is a dollar somebody inside an agency holding company needs you to spend.
The $15 trillion number is doing too much work
Every agentic commerce deck in B2B cites the same Gartner prediction. By 2028, 90% of B2B buying will be AI-agent intermediated, pushing more than $15 trillion of B2B spend through automated exchanges.
That’s the cudgel. Get ready, or get bypassed.
Read the underlying language. “Intermediated” is doing a lot of work in that sentence. Gartner is counting any deal where a buyer used a chatbot during research, where a procurement tool re-ranked a vendor list, where an internal copilot summarized an RFP. That is not “an agent bought the software.”
When you ask buyers what they’re actually doing, the picture collapses.
Forrester’s 2026 State of Business Buying surveyed nearly 18,000 global buyers. Yes, 94% used AI somewhere in the purchasing process. The next sentence in the report is the one no vendor deck quotes: buyers compensate for AI’s limitations by validating what they find against other sources. They are not delegating. They are double-checking. 🤖
G2’s 2026 software buyer data shows 71% of B2B software buyers now start research with an AI chatbot. Start. Not finish. Not authorize. Not pay.
The Averi analysis of 680 million procurement-related citations: 73% of B2B buyers use AI tools like ChatGPT and Perplexity for research. Same word. Research.
The gap between “I asked ChatGPT what the best CRM is” and “my agent autonomously bought a CRM for me” is the entire chasm vendors are pretending doesn’t exist.
What buyers actually still do
Here is what hasn’t changed in the last 24 months, despite every keynote suggesting otherwise.
The average B2B buying committee is between 5 and 13 stakeholders, depending on whose research you read. Influ2’s 2026 enterprise buying survey: 32% of buyers say a senior executive’s opinion carries the most weight, and another 32% say it’s a group decision. Almost nobody — a rounding error — said an AI agent decides.
74% of buyer teams experience unhealthy internal conflict during the decision.
86% of B2B purchases stall during the buying process. That number has been roughly stable since Forrester started publishing it.
This is the actual bottleneck. Not “the buyer doesn’t have machine-readable product data.” Not “your schema isn’t agent-friendly.” Not “you haven’t joined the OpenAI advertising pilot.”
Eleven humans in different functions can’t agree on whether the risk is worth the budget. ✋
No agent solves that. No agent has been invited to that meeting. No agent gets fired when the implementation flops in Q3.
The vendor receipt
So why is everyone selling you the agent future this hard?
Look at who’s buying.
Publicis paid a 29.8% premium over LiveRamp’s prior close to lock in “data co-creation for smarter agents.” It is a $2.2 billion bet that an agency holding company can sell agentic-AI services to brands faster than the next four quarters of earnings calls demand. The agentic story is what’s keeping their forward multiples intact.
Salesforce closed 29,000 Agentforce deals in 15 months and grew Agentforce ARR 169% year-over-year to $800 million by the end of fiscal 2026. The proxy statement now formally measures executive comp on “Agentic Work Units.” Their entire investor narrative depends on you believing the agents are coming for your stack.
Google announced Ask Advisor at Marketing Live this Tuesday — an AI agent layer inside Ads, Analytics, and Merchant Center. The pitch: your ads become the best answer to agent queries. Buy more inventory. Train Gemini on more of your first-party data.
Criteo’s pilot with OpenAI scaled to 1,000+ brands in roughly ten weeks, with case studies showing AI-attributed conversions running near 2x. Every adtech vendor is now racing to slap a “ChatGPT-ready” label on its inventory.
This is a tooling buildout pretending to be a buyer behavior shift. The pickaxes are real. The gold rush is not. 📉
If you doubt me, read Forrester’s own 2026 predictions. The same firm that publishes the “94% of buyers use AI” number also projects that B2B companies will lose more than $10 billion in 2026 from ungoverned use of generative AI. And it forecasts that only 1 in 5 B2B sellers — 20%, not 90% — will be compelled to respond to AI-powered buyer agents with agent-controlled counteroffers this year.
20% in 2026. 90% in 2028. Connect those two dots without a vendor’s hand on the pencil.
Even Bessemer, deep in the agentic commerce thesis, conceded in a note published this Friday that the “seamless delegation promise” is not here yet. They’re in the trade and they’re still hedging.
Meanwhile, McKinsey’s 2026 State of Marketing report says CMOs rank brand as their top priority for the second straight year. AI search ranks 17th out of 20. 94% of CMOs report no meaningful progress integrating AI into marketing operations.
Those are the people the agency holding companies are about to sell $2.2 billion of agentic-AI services to. 🤡
What to actually do this quarter
If your CRO walks in next week with a slide called “agent readiness,” ask one question: how many buyer-side agents bought from us last quarter?
The answer is almost certainly zero.
Now ask how many deals stalled inside the buying committee. The answer is most of them.
That’s your roadmap. Pull budget away from the agent-future line items and put it back on the work that fixes consensus.
Build a separate content asset for every persona on the committee. Not “a CFO-friendly version of the same blog post.” A different artifact entirely. ROI math for finance with sources they can verify in twenty minutes. A risk and compliance brief for security and legal — the people who actually kill deals, not the ones who sign them. An implementation timeline for IT that doesn’t read like a brochure. A change-management note for the operations lead who has to absorb the disruption after go-live.
When the deal stalls — and Forrester says 86% of them do — your champion needs ammunition they can hand to the dissenter without setting up a follow-up call you’ll never get. Most B2B websites still don’t have that ammunition. Their resources page is six pillar blog posts and a gated ROI calculator. That’s an MQL trap, not a deal accelerator.
Audit your last ten lost deals. Find out which stakeholder killed each one. Build content specifically for that role on the next ten. Cheaper than another Agentforce SKU and an order of magnitude more likely to move pipeline.
Stop optimizing your product feed for Gemini until somebody on the buyer side actually asks you to.
The agentic commerce future will arrive. Some version of it. In some industries. On a timeline nobody can predict from a podium in Dublin or a press release dated Sunday night.
It is not arriving this quarter. It is not arriving next quarter. And it is not arriving in time to justify the line items your CFO is currently squinting at.
What’s arriving is a wave of vendor pressure designed to make you feel slow. Publicis needs you to feel slow. Salesforce needs you to feel slow. Every agency that lost an SEO retainer two years ago and rebranded as “agentic AI consulting” needs you to feel very slow.
The buyer isn’t sending an agent yet. The buyer is sending eleven humans who can’t agree on a renewal cycle.
Sell to those eleven. The agent can wait.
