Your Next Buyer Is an AI Agent: Selling in the Agentic Era

A few years ago, if you sold B2B software, you knew exactly who you were selling to: a person. A head of sales who picked up the phone, a finance director who haggled on seats, a champion who went to bat for you internally. The buying committee was made of humans, and humans respond to rapport, proof, and persistence.

That assumption is quietly dying. In 2021, Maersk — the world's largest container shipping company, $54 billion in revenue — deployed AI agents from a startup called Pactum to negotiate freight lane contracts with its carriers. Not to support human negotiators: to run the whole thing. Each agent reached out to carriers, ran multiple rounds of negotiation on pricing, routes, and payment terms, and closed the deal. No human touched any single transaction. The result: a 96% agreement rate, and in controlled trials against human negotiators, the agents secured rates 22% lower on identical shipping lanes.

That is not a futuristic scenario. That is a case study from 2021, and the buyer-side infrastructure has only gotten smarter since. If you are a seller, a founder, or a GTM team and your commercial playbook still assumes a human on the other end of the table, this article is for you. We're going to walk through the actual numbers behind agentic buying, the three pillars of your business model that break first, and the practical playbook for surviving — and winning in — a world where your buyer is a bot.

The shift is real, and it's measurable

Let's start with the number every seller should have tattooed on their dashboard. In October 2025, Gartner unveiled its top strategic predictions for 2026. Prediction one, straight from the press release: by 2028, 90% of B2B buying will be AI-agent intermediated, pushing over $15 trillion of B2B spend through AI agent exchanges. Roughly half of US GDP moving through machines transacting with machines.

You can argue with the date, but you can't argue with the direction. And the leading indicators are already here:

• Gartner, March 2026 (survey of 646 B2B buyers, Aug–Sep 2025): 67% of B2B buyers now prefer a rep-free experience, up from 61% a year earlier, and 45% report using AI during a recent purchase.
• Gartner, Aug 2025: by the end of 2026, 40% of enterprise applications will embed task-specific AI agents — up from under 5% a year earlier.
• G2, April 2026 (survey of 1,000+ B2B software buyers): AI chatbots now top the list of sources influencing vendor shortlists — ahead of software review sites and vendor websites. 51% of buyers now start their research with AI chatbots, up from 29% the year before. And 85% of buyers rate a vendor more highly when an AI answer engine includes them in a response.
• Deloitte, 2026 (1,000+ US suppliers and buyers): nearly 40% of B2B buyers already use agentic AI in purchasing — evaluating products, configuring orders, reviewing contracts, benchmarking prices. Supplier adoption is trailing: only 24% use agents in the sales process (though 67% plan to).

Notice the asymmetry. Buyers are ahead of sellers. They have already outsourced the research, the shortlisting, and increasingly the negotiation to software. Most sellers are still optimizing their homepage for humans.

What changes first: the research phase is now owned by machines

The most practical way to understand this shift is to watch what happens before a buyer ever talks to you.

HubSpot's Director of Product Development, Beeri Amiel, put it bluntly in April 2026: "By the time they're getting to your website, they're already much further down the funnel. All the selling was done by the answer engine." Sam Senior, founder of TestBox, reports that his enterprise seller customers now observe that 70–80% of the buyer's decision is already made before they ever speak to a salesperson.

This is consistent with Forrester's long-running finding that 68% of B2B buyers already have a front-runner vendor in mind at the start of the buying process. The difference is that, today, that front-runner is chosen in the background by an agent reading your website, your docs, your pricing page, your G2 reviews — and synthesizing them into a shortlist. The human buyer sees the output, not the process.

This is why the previous article in this series on Generative Engine Optimization (GEO) matters: if an AI answer engine doesn't include you, you're effectively invisible at the exact moment the shortlist is formed. But GEO is only half the story. Once the agent has you on the shortlist, it then has to evaluate you — and that's where most B2B commercial infrastructure fails.

The three pillars that break first

Every traditional B2B go-to-market model rests on three assumptions. All three are being quietly dismantled by agent buyers.

1. Per-seat pricing breaks when the "user" is a process

Per-seat subscription pricing assumes a human user who opens discrete sessions. An AI procurement agent doesn't work that way — it spawns parallel sub-processes, runs 24/7 across time zones, and consumes your product in bursts that don't map to any seat count.

Consultancy Kearney estimates that AI procurement agents could erode up to 500 basis points of EBIT for distributors by commoditizing supplier selection and compressing average selling prices by about 8%. That's a P&L consequence finance teams can measure, not an abstract pricing debate.

The sellers who see this coming are already restructuring. Sierra (the a16z-backed AI customer service company) abandoned seat-based pricing entirely for outcome-based pricing — clients pay when the software delivers a specific, high-value result. Intercom did the same with its Fin AI agent: $0.99 per successfully resolved conversation, unresolved interactions free. As Intercom's president Archana Agrawal put it: "Customers didn't want to pay for activity, and so we get paid when our customers have that positive outcome."

The lesson: agents evaluate vendors on quantifiable value per result. If your price is a flat per-seat number and the agent can't derive the value-per-outcome, you lose the comparison.

2. Relationship-driven sales breaks when the counterparty can't be charmed

Every classic sales playbook assumes a human counterpart with career stakes in the relationship, a memory of prior interactions, and susceptibility to persuasion. Agents have none of that. They evaluate purely on metrics and pass the results to a human principal.

The SUEZ example shows how fast this compounds. SUEZ UK, part of the €19 billion SUEZ Group, deployed Pactum's agents and reached 2,000 additional suppliers within two months, achieving average potential savings of 2.5% and cost reductions of 15% purely through competitive purchasing pressure. Those vendors were negotiating against an automated counterpart that never got tired, never got flattered, and always compared offers on the numbers.

This also corrodes customer success. Clari Labs analyzed 10 million opportunities from 121 major global enterprises (Jan 2023–Dec 2024) and found the average contract value fell 50% year-over-year, while the average expansion deal cycle grew from 92 to 125 days. Their read: buyers now require verified evidence of value before approving upgrades. When an agent evaluates your renewal, it computes ROI from API usage logs, cross-references competitor pricing programmatically, and presents the delta. A great relationship with your CSM no longer saves you if the machine-readable performance data isn't there. By the time a CSM initiates the renewal conversation, the agent has already finished its evaluation.

3. Commercial authorization breaks when the buyer has no human hand

Here's the least-discussed bottleneck: an AI agent cannot finalize a transaction or commit budget without verifiable proof of its authority. If your checkout flow requires a human to click "accept," the autonomous workflow hits a hard stop — and the agent simply moves to a vendor whose systems let it finish.

This is the "agentic identity" problem, and it's huge. Entro Labs' H1 2025 report found non-human identities now outnumber human identities 144 to 1 across enterprises. In financial services alone, a16z affiliate Sean Neville notes non-human identities outnumber human employees 96 to 1. Sellers need to redesign checkout and authorization flows so an agent can programmatically verify its spending limits and legal authority — otherwise you're structurally cut out of the agent-mediated deals.

The seller-side paradox: more agents, less productivity

Here's the uncomfortable twist. While buyers are automating, most sellers' own AI adoption is not paying off.

Gartner's November 2025 sales research predicts AI agents will outnumber human sellers tenfold by 2028 — yet fewer than 40% of sellers will report that AI agents improved their productivity. A July 2026 Gartner release doubles down: CSOs who overhaul their data, automation, and user experience will be five times more likely to gain ROI from AI than those taking quick fixes. Meanwhile, 60% of CSOs say their revenue number is largely driven by elements outside their control.

Gartner analyst Dan Gottlieb's diagnosis is precise: "Without the right data foundation, workflow integration and seller experience, CSOs risk creating agent sprawl, with more digital activity, but little improvement in seller impact." If your systems are fragmented, agents scale the fragmentation. The answer isn't more AI tools bolted onto a broken process — it's rebuilding the commercial infrastructure so AI can actually operate on it. That's exactly the same conclusion the buyer-side data points to.

The playbook: how to sell when the buyer is an agent

Enough doom. Here's the practical side, distilled from the sellers that are actually winning. None of this requires a massive budget — most of it is content, structure, and process discipline.

1. Build a machine-readable product surface

An agent shortlists vendors based only on the structured information it can consume. If your product specs exist only in a glossy PDF or a marketing page, you risk being skipped. Publish a single, canonical, public, machine-readable spec page: SKUs, integrations, API endpoints, security certifications (SOC 2, ISO 27001, HIPAA, PCI DSS), supported regions, deployment options, implementation time, SLAs. Mark it up with JSON-LD Product and SoftwareApplication schema so crawlers and agents can parse it cleanly.

There are emerging standards here worth watching: Google's Agent2Agent (A2A) protocol uses "Agent Cards" — structured JSON capability documents — so agents can discover and assess vendor capabilities. The Universal Commerce Protocol (UCP) aims to make product info, pricing, availability, and checkout readable by AI systems, with backing from major commerce and payments companies. The vendors without a machine-readable capability profile are becoming invisible to agent-driven shortlisting.

2. Atomize your proof

Agents love discrete, verifiable claims. Take every customer success story and break it into 8–12 individual, citable facts: named customer (with permission), specific metric, date, source. A single "95% of customers see ROI" claim is weak; a machine-parseable list of "Customer X, +34% pipeline in 90 days, verified Q3 2025" is ammunition. One consistent description of what you are must appear across your site, your review profiles, and directories — entity consistency is how agents connect the dots.

3. Publish your pricing (or at least the shape of it)

You don't have to publish one fixed number. Publish ranges, tiers, or an interactive pricing calculator. Agents compare offers; if your pricing is hidden behind "Contact Sales," an agent evaluating vendors has nothing to compare and drops you. Transparency is increasingly a requirement for being shortlisted, not a concession.

4. Build a public, machine-readable trust center

Security and compliance questions are the first thing an agent checks. Maintain a public trust center: current SOC 2 Type II status, ISO certifications, GDPR posture, EU AI Act assessment, DPA template, sub-processor list, pen-test summary. Update it continuously. If an agent can't verify your compliance posture, it can't recommend you — the risk is simply too high.

5. Publish honest comparison content

Agents lean heavily on "X vs Y" content during shortlist construction. Publish well-structured comparison pages for the three to five competitors you most often get paired with. Real tables: capability, pricing model, integrations, security certifications, deployment time, support model. This is counterintuitive — most sellers fear giving competitors oxygen — but comparison pages are where agents make decisions, and an honest table beats being absent.

6. Build a structured RFP response engine

More procurement is moving to structured RFPs evaluated by agents. An internal AI-powered response engine that ingests a structured RFP, matches every line item to your canonical answer library, drafts a response, and routes exceptions to humans will win deals that a human team writing from scratch simply can't keep up with.

7. Train your humans to speak agent

Your reps will increasingly talk to buyers who have an AI co-pilot listening, summarizing, and scoring. Teach them to lead with verifiable outcomes, then context — high-density, well-cited statements that survive summarization. What gets clipped into the AI's summary is what wins; vague positioning copy gets discarded. Also track which AI tools your buyers used: add UTM parameters, AI-source detection, and explicit form fields ("How did you first hear about us?" with ChatGPT, Perplexity, Claude, Gemini, and Microsoft Copilot as options) so you can attribute and optimize.

8. Audit where you stand right now

Do this this week: query ChatGPT, Perplexity, Claude, and Google AI Overviews for your category keywords and your buyers' most common questions. Do you appear in the generated answers? Do competitors appear instead? That's your share-of-model-voice baseline. Then rewrite your most important content for extraction — named sources, concrete numbers, clear process steps, direct Q&A — because that's what gets surfaced. This is the GEO discipline from the previous article, applied to the whole commercial surface.

What this means for founders and small teams

If you're an indie founder or a small GTM team, this shift is actually good news in one important way: agent-mediated buying flattens the playing field. A giant enterprise's advantage was the relationship-driven sales machine — the armies of SDRs, the dinners, the multi-touch campaigns. Agents don't care about any of that. They care about structured data, verifiable outcomes, and clear pricing — all things a small, technical team can produce faster than a big company can restructure.

Your asymmetric advantage: you can make your product machine-readable in a week. A public spec page, a transparent pricing page, an atomized case-study library, a trust center, three comparison pages. That's a 90-day project for a small team, and it's precisely where the enterprise incumbents are slowest to move.

One warning: don't over-invest in building your own buyer agent before you've made your own surface agent-readable. The order matters. First become legible to the agents that are already out there buying — then, if it makes sense, deploy your own agents to sell against them.

The bottom line

The buyer-side transformation is not coming; it's here, and it's lopsided. Buyers are deploying agents faster than sellers are preparing for them. The sellers who redesign their pricing, their proof, and their infrastructure for machine readers will hold a structural advantage that slower competitors will never catch — because every month the agents get better at choosing, and every month a vendor stays invisible to them, the gap widens.

Gartner's prediction is a two-year runway. The question isn't whether your next buyer will be an AI agent. It's whether you'll be visible, legible, and buyable when it arrives.

Disclosure: This article was researched and written with AI assistance. All statistics are drawn from the named public sources listed below; where figures conflict across sources, methodology is noted. No metrics in this article are fabricated.

References

1. Gartner, "Gartner Unveils Top Predictions for IT Organizations and Users in 2026 and Beyond," Oct 21, 2025 — 90% of B2B buying AI-agent intermediated by 2028; $15T+ through agent exchanges.
2. Gartner, "Gartner Sales Survey Finds 67% of B2B Buyers Prefer a Rep-Free Experience," Mar 9, 2026 (survey of 646 buyers).
3. Gartner, "Gartner Predicts AI Agents Will Outnumber Sellers 10 to 1 by 2028…," Jul 28, 2026 (survey of 210 CSOs).
4. Gartner press release, "AI Agents Will Outnumber Sellers 10x by 2028…," Nov 18, 2025 (Melissa Hilbert).
5. G2, "The Answer Economy" research, Apr 2026 (1,000+ B2B software buyers) — as reported by O'Reilly Radar.
6. Deloitte, "Agentic commerce: The future of B2B commerce," Jun 26, 2026 (research of 1,000+ US suppliers and buyers).
7. O'Reilly Radar, "When Your Buyer Is an AI Agent," Aug 19, 2026 — Pactum/Maersk and SUEZ case studies; Sierra and Intercom outcome pricing; Clari Labs benchmark; Entro Labs NHI report; a16z (Catena Labs) non-human identity figures.
8. McKinsey, "The State of AI" global survey, Nov 2025 (1,993 respondents) — 62% experimenting with AI agents.
9. McKinsey, "Redefining procurement performance in the era of agentic AI," Feb 2026 — chemicals and telco procurement pilot results.
10. Cloudflare, "Crawlers click: AI bots training" blog, Jul 2025 — AI bot crawling +24% YoY; CEO expects automated traffic to pass human by 2027.
11. Kearney (via Distribution Strategy), Oct 2025 — AI procurement agents could erode ~500bps EBIT for distributors; ~8% ASP compression.
12. Forrester, Buyers' Journey Survey — 68% of B2B buyers have a front-runner vendor in mind at start.
13. HubSpot (Beeri Amiel) and TestBox (Sam Senior) commentary, Apr 2026 — buyers decide 70–80% before contacting sales.
14. Salesforce, State of Sales 2026 — 54% of sellers have used AI agents; high performers 1.7x more likely to use prospecting agents.
15. Gartner IT Symposium/Xpo 2025 Orlando Day 2 Highlights — "Design processes for AI agents, not humans."

Suggested visuals (placeholders)

• Cover/OG image: an abstract diagram of two agent nodes exchanging structured data packets, "agent-to-agent" motif.
• Mid-article infographic: timeline of buyer-side agent adoption (2021 Maersk → 2025 67% rep-free → 2028 90% intermediated).
• Table visual: "per-seat vs outcome-based pricing" comparison for the pricing section.
• Checklist graphic: the 8-step agent-ready playbook.