AI in the B2B buying journey

# AI in the B2B Buying Journey: Where It Helps, Where It Doesn't Yet, and What Suppliers Should Prepare *Part of the B2B Commerce Corner series.* Recent research makes one thing clear: B2B buyers have adopted AI quickly. Forrester reports that 94% of business buyers use AI in their buying process, and that buyers now name generative AI as their most meaningful information source twice as often as any other source. Gartner adds an important counterpoint: 69% of B2B buyers prefer to validate AI-generated insights with a sales rep. Both findings are useful. This article adds the context suppliers need to act on them: what kind of purchase we are talking about, what AI actually does in that purchase, and what it asks from your content and your systems. ## What the research says Gartner's survey of 645 B2B buyers found that buyers used an average of seven information sources in a recent purchase, and that 45% used GenAI, mainly to gather information on vendors and products. Buyers want self-service: 67% prefer a rep-free experience and 70% prefer a fully digital one. Yet reps remain the most important source at key moments: defining the problem, selecting a preferred supplier, building internal support and finalizing the purchase. Forrester adds that 61% of business buyers use private AI tools provided by their own organization, and that buyers say these tools help them consider more vendors in less time and make better decisions. The picture is consistent. AI has become the starting point of the research, and people remain part of the decision. ## Start with the purchase, not the percentage What the headline figures don't show is the type of purchase. In B2B, that makes all the difference. We find it useful to look at buying through order archetypes rather than industries, because ordering behavior cuts across sectors: **Replenishment and contract call-off.** The buyer already has a supplier, a price agreement and a routine. They don't ask an AI where to buy. They reorder through the channel they already use. AI plays a small role here today. **Incident-driven parts.** A machine is down and the part is needed now. If the regular supplier has it, the buyer calls or orders there. If the original supplier is gone or the part is discontinued, the search starts, and AI is increasingly where it starts. **New sourcing.** A new product category, a new supplier, or a contract up for renewal. This is where AI does most of its work. **Seasonal pre-orders and configured products.** These are mixed. AI helps with orientation and specification, while configuration, volumes and terms still go through people. So the question for a supplier isn't "do our buyers use AI?" It's "how much of our revenue depends on buyers who are still looking, and how visible are we to them?" ## What AI does: the first sift When a buyer has a need but no supplier, AI takes over the most time-consuming part of the early work. That used to mean searching, opening websites, downloading datasheets and comparing specs. Now the buyer describes the need and gets a shortlist. After that, people take over. Gartner's finding that buyers validate AI output with reps fits this pattern. Notably, buyers expect misleading information from AI (51%) and from reps (49%) about equally often. The rep's value in this phase isn't having better facts. It's accountability and context. That makes the first sift more important than it looks. A supplier that isn't on the AI's shortlist never reaches the stage where a rep can make a difference. ## Why GenAI beats traditional search for B2B B2B questions are combinations of requirements: a material, an operating condition, a certification, a delivery region, a timeline. Search engines return links and leave the comparing to the buyer. GenAI combines the requirements, compares the options and narrows the answer as the buyer adds constraints. The private AI tools Forrester describes add something search never had: the buyer's internal context. A buyer's company assistant can combine public supplier information with internal specifications, requirement documents and purchase history. That makes the shortlist more relevant for the buyer, and more demanding for the supplier. ## Is your content prepared for it? AI can only recommend what it can read and understand. On many B2B websites, the most important information is hard for a machine to use. Common problems: - Specifications only available in PDF datasheets - Product names that are internal article codes rather than descriptions - No content about applications: what the product is for and when to choose it - Certifications and compliance information that are hard to find - "Contact us" in places where standard information, like typical lead times or delivery regions, could be published - Relationships between products, parts and accessories that only exist in the ERP or in people's heads Fixing these is not a separate "AI project." It's product data done properly. The same structured product information that makes you visible to AI also makes a customer portal usable for the buyers you already have. ## What to open up and what to keep behind the login Not everything should be public. Contract prices, customer-specific assortments, account-level availability and order history belong behind a login. A useful rule of thumb: **Open and structured:** everything that helps a new buyer decide whether you're a fit. Products, specifications, applications, certifications, markets served and standard terms. **Authenticated:** everything that depends on the relationship. Prices, assortments, availability per account, order status and history. Many suppliers gate more than necessary. If you don't publish the first category, AI will still answer the buyer's question, using someone else's information or an estimate. ## Where agents come in AI assistants answer questions. AI agents take actions: requesting quotes, comparing offers, checking availability and eventually placing routine orders on the buyer's behalf. That brings AI into existing supplier relationships, the part of B2B buying it barely touches today. But agents need structured, authenticated access to exactly the information that stays private: customer-specific prices, availability, order status and order history. An agent can't reorder from an inbox or a PDF price list. This is still early. Most B2B purchases today don't involve agents. But the requirements are the same ones a good self-service setup already has: reliable data, customer-specific terms and clear order processes. ## What to do now 1. **Map your revenue by purchase type.** Know how much of your business comes from new sourcing and how much from repeat and contracted ordering. 2. **Run the test.** Ask AI tools the questions your recent new customers asked before they found you, and see who gets recommended. 3. **Fix your product data.** Publish structured specifications, descriptive names and application content rather than PDFs and article codes. 4. **Draw the public/private line deliberately.** Open up what helps a stranger evaluate you, and keep relationship data authenticated. 5. **Make relationship data accessible in a structured way.** Customer-specific prices, availability and order status should be available to logged-in buyers today, and to their systems tomorrow. --- *Written by [Falco van der Maden](https://www.linkedin.com/in/falcovandermaden), founder of Vendordesk, a B2B ordering portal and operations platform for manufacturers, wholesalers and distributors. Falco has worked on B2B commerce with manufacturers and wholesalers since 2005, and sees both sides of this topic every day: how new buyers find suppliers, and how existing customers keep ordering from them.* *Sources: [Gartner, May 20, 2026](https://www.gartner.com/en/newsroom/press-releases/2026-05-20-gartner-survey-finds-sixty-nine-percent-of-b-two-b-buyers-turn-to-sales-reps-to-validate-ai-generated-insights) · [Forrester, B2B Buyers Make Zero-Click Number One](https://www.forrester.com/blogs/b2b_buyers_make_zero_click_buying_number_one/)*