McKinsey Says AI Agents Now Run Overnight Vendor Analysis. The Deal Was Decided Before the First Call Was Booked.
McKinsey's February 2026 procurement report describes AI agents running overnight vendor evaluations end-to-end. Gartner projects 90% of B2B buying flows through AI agent exchanges by 2028. Most brands are still optimizing for the search layer. The procurement layer is already running.
McKinsey published a procurement report in February that describes something most B2B revenue teams haven't fully processed. An AI agent that "analyzes supplier bids overnight." That "tracks market indices in real time." That "operates end to end: from identifying a strategic opportunity to sourcing suppliers, to preparing commercial strategies, to tracking performance post-award." The report is called Redefining Procurement Performance in the Era of Agentic AI. It's not forecasting a capability that might emerge. It's describing deployments already underway at enterprise organizations — pharmaceutical companies, technology firms, manufacturers — that have decided the procurement function is where agentic AI delivers the fastest return.
This is a different AI moment than the one most marketers are preparing for.
The search layer versus the procurement layer
The dominant framing of AI visibility focuses on search: ChatGPT, Perplexity, Google AI Overviews. A prospect queries a category. You want to appear in the answer. That framing is real and the implications are significant — but it locates the AI moment at the top of the funnel, which is where the stakes are comparatively low. The prospect is still exploring. They haven't committed to a vendor set. There's still time to make an impression, to show up in a follow-up query, to get introduced through another channel.
What McKinsey is describing sits somewhere entirely different: not at the start of the consideration process, but before the formal conversation exists at all. When a procurement team at an enterprise organization deploys an AI agent to evaluate the vendor landscape in their category, they aren't asking it to suggest options for future exploration. They are asking it to build the brief. The agent runs overnight. It synthesizes supplier intelligence from market data, published reports, editorial coverage, and the kind of third-party sources that enterprise procurement teams have historically relied on to make initial credibility judgments. By the time a procurement manager looks at the output, the analysis is complete. Some vendors made the brief. Most didn't.
The meeting request you receive — if you receive one — comes after the AI already ran your evaluation.
The scale of what's coming
Gartner's "Top Strategic Predictions for 2026 and Beyond" is more specific than most people have registered. The forecast: by 2028, 90% of B2B buying will be AI agent intermediated, channeling more than $15 trillion in spend through automated agent exchanges. That's not 90% of marketing touch points or 90% of research queries — it's 90% of the actual purchasing decision flow. The economic infrastructure of B2B commerce is being rebuilt around AI mediation, and enterprise procurement is where the structural change is most advanced because procurement is where the business case for agents is clearest. AI agents don't get tired. They don't miss signals in the data. They can analyze a hundred vendor profiles overnight and deliver a ranked shortlist by 7 AM. The ROI is immediate and measurable, which is why McKinsey found leading companies already seeing "outsized results" — one tech company used a "linked set of AI agents to rebuild its strategy for sourcing external services," with one agent integrating spend and market data to generate real-time savings insights while another simulated demand scenarios for hedging.
Fortune Business Insights projects the global agentic AI market will grow from $9.14 billion in 2026 to $139 billion by 2034. The procurement function isn't waiting for that market to mature. It's already deploying.
The gap in current AI visibility strategy
Most of the work happening under the AI visibility umbrella treats this as a marketing and content problem. Optimize your website for AI crawlers. Structure your product data so agents can parse it. Get cited in AI overviews for your target queries. These are real moves that belong in any serious strategy.
But they're moves designed for the search layer — the AI that responds to explicit queries from individual users who are still building their understanding of a problem space. What McKinsey describes operates at a different level and evaluates vendors by a different standard.
A procurement agent isn't assessing your brand the way a researcher uses Perplexity to explore options. It's assessing your brand the way a senior procurement professional would — which means heavy weighting on whether the vendor has a verifiable, independent track record in the market. Not website copy. Not schema markup. Not the product description you crafted for AI readability. The kind of record you can cross-reference from sources the procurement function has always trusted: independent editorial coverage in respected publications, analyst reports, earned media that demonstrates your firm's standing in the category.
The procurement agent applies the same trust hierarchy the procurement team itself applies. It just does it faster, at a larger scale, without breaks.
What the overnight analysis actually draws from
The publications enterprise procurement teams have relied on for decades — TechCrunch, Forbes, Bloomberg, Harvard Business Review, category-specific trade journals — are the same publications AI systems weight most heavily when evaluating vendor credibility. This reflects how these systems were trained: they learned to assess sources the same way well-read humans did, and those humans consistently assigned the highest trust to independent editorial coverage in respected outlets. Product-page copy sits near the bottom of that hierarchy. Earned media in authoritative publications sits at the top.
MIT Sloan Management Review documented the practical consequence in January: even market-leading brands with the largest search investments found themselves absent from AI recommendations in their own category. "That was a wake-up call," an executive told the researchers after discovering a smaller competitor dominated the AI answers. "We were shocked." The mechanism behind that shock is the same one operating in AI procurement evaluation — the AI learned what humans learned to trust, and is now applying those trust judgments at machine speed and enterprise scale.
A procurement agent building an overnight vendor brief on your category will construct each vendor profile from its editorial record. If your record is strong — consistent coverage in publications your buyers have always read — you'll be represented accurately and favorably. If your only editorial presence is a press release from 2023 and a listing in a category directory, the agent constructs your profile from thin material and the brief reflects that. You may not be eliminated explicitly. You just won't be the answer to the question that gets the deal.
The window that's closing
The AI procurement deployment McKinsey describes isn't a 2028 event. It's happening in 2026 at organizations that have already built the business case. The Gartner $15 trillion forecast describes where the infrastructure ends up, not where it starts. By the time most B2B brands treat the procurement layer as seriously as they treat the search layer, the briefing documents will have been running long enough to establish default answers that are hard to displace.
The mechanism that determines those default answers is the same mechanism that has always shaped procurement trust at the executive level: earned media placements in publications the category trusts, accumulated over time, consistent enough that an AI synthesis of your brand's track record produces a clear and favorable picture. This is what Machine Relations addresses — not the data schema, but the editorial record that AI procurement agents draw from when they decide which vendors belong in the brief. PR's mechanism always worked. The reader running the overnight analysis just moved from human to machine.
Run the audit to see where your brand stands in that record before the next overnight brief gets built without you in it.
Related Reading
- AI Visibility for Consumer Brands: The 2026 Earned Media Playbook
- AI Visibility for Fashion: The 2026 Earned Media Playbook
Related reading