In B2B, Agents Narrow The Field—But The Human Still Buys

Rob Green is the chief digital officer of Insight Enterprises, a Fortune 500 Solutions Integrator.

gettyOnline B2C consumer behavior is often compared to spearfishing: targeting a specific product for optimal pricing, fulfillment and customer service. Consumers know what they want, find it, possibly add unplanned items to their carts based on cross-sell recommendations and hit the buy button. It’s the opposite of casting a wide net.

However, that’s consumer behavior. John Dawes’ 95:5 rule holds that only 5% of B2B buyers are spearfishing for specific solutions in a given quarter. They buy in bulk for a specific use case or a specific persona, with a range of purchase criteria.

Both buyers are about to offload that work to an agent, and the difference in what they’re offloading matters.

With AI-enabled buying entering the mainstream, will B2C and B2B agentic buying trends mirror e-commerce trends of the past 20 years? It’s hard to say because most everything we’ve heard about agentic buying is biased toward B2C buyers.

Meta’s Muse, launched in September, checks out for consumers at major retailers, but Amazon blocked it within two weeks. Every headline is about who gets to buy for the shopper. McKinsey’s own agentic commerce report targets consumers and merchants, but B2B only appears once in a parenthetical aside as a “significant” category its figures omit.

I don’t think it’s safe to assume B2B agentic buying will lag B2C, which means the B2B-specific effects of AI need your attention. It’s now in the top five channels B2B buyers rely on to find, research, short-list and evaluate suppliers. How do we adjust to a B2B buying journey that moves faster than the humans inside it?

A 2026 G2 report found that 51% of surveyed B2B software buyers began the purchasing process with AI rather than a traditional search engine, while a 2026 Gartner, Inc. survey found that two-thirds of responding B2B buyers preferred a rep-free experience outright.

Those two numbers reconcile once you understand which actions B2B buyers are asking AI to complete. According to IDC, “buyers now rely on these systems to narrow the field before human evaluation even begins.” Narrowing isn’t buying. B2B buyers are asking agents to do their research and deliver informed recommendations. They’re not asking AI to complete the purchase process (at least not yet).

B2B buyers seek informed, reliable and accurate responses to their specific use cases. We had education buyers who came to us and asked, “What’s the best laptop for a K-12 school system with 5,000 students?” The buyers expect an instantaneous answer that demonstrates judgment, expertise and an awareness of how the category has moved since they made their last sizable order of laptops.

The category keeps moving. What used to be a straightforward bulk laptop purchase is now significantly more complex. Am I going to move AI workloads onto these machines? If so, which laptop should I buy?

Notice that the role an AI agent plays in a B2B buying journey isn’t the one a salesperson plays. It’s a purchase advisor built to decomplexify. Its job is helping a buyer stay up to speed in a category that keeps moving faster than anyone tracking it part-time can follow, and buyers are listening. The G2 report found that 69% of respondents said they chose a different vendor than the one they’d planned on based on what the chatbot told them.

What’s true of your GPS is also true of AI: The guidance is only as reliable as the data supporting it. One of the key insights we had when developing our procurement agent was that product data fidelity needs to be accurate and reliable. Absent precise product data, an agent makes poorly informed recommendations while sounding every bit as certain as a good one.

B2B buyers can’t afford to make poorly informed decisions. Forrester surveyed nearly 18,000 of them and found that answer engines “often deliver incomplete or unreliable information, creating mistrust” and that “buyers compensate for this by seeking validation from trusted sources.” According to McKinsey, inconsistent information from a supplier’s own teams is now the number one reason they switch suppliers, ahead of the inability to reach a knowledgeable representative.

In ideal conditions, B2C AI agents find the best option available, whereas a B2B AI agent identifies the best solution a B2B buyer is seeking within their buying constraints or guidelines. Buying guardrails include approved company standards, budget constraints, approval workflows and timing.

Procurement sets those rules up front, and the agent works within these constraints to support buying decisions. OpenAI’s dots, launched September 29, ship with rules for when the agent must ask for approval and when it may act, and its first enterprise pilots include procurement. Procurement professionals, which Forrester found to be “decision-makers in 53% of business buying cycles,” are that human in the loop.

1. Product data must be accurate and reliable. Audit your product data before launch day.

2. Encode buying rules as guardrails. Product data informs the agents of buying options. Ask yourself what the agent needs to know to support the buying decision.

3. Clarify the agent’s role for salespeople. If a sales team views an agent as a threat to their livelihood, they’ll be blockers of agentic buying, not promoters.

4. Track the questions it can’t answer. Keeping a running list of questions that go beyond the scope of what the agent was built to handle functions as an honest product roadmap.

The sales wisdom that says customers buy from people they “know, like and trust” hasn’t changed. The only difference now is that the advisor can be an agent, and it has the same trust to earn. Get the data right and the humans in the right place, and I believe B2B buyers will also come to know, like and trust your agent.​

(Full disclosure: Insight Enterprises has commercial relationships with Amazon Web Services, Meta and OpenAI.)

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?