The Self-Directed, AI-Assisted Buyer

What the 2025–2026 evidence says

Three recent studies describe the same buyer from different angles.

Gartner surveyed 632 B2B buyers in August and September 2024 and reported in June 2025 that 61% prefer an overall rep-free buying experience. Buyers leaned toward self-service for general information but still sought seller input on tasks such as assessing whether a product fits their organization's needs.

Forrester's State of Business Buying 2026, based on nearly 18,000 buyers, found that 94% use AI somewhere in their buying process and that generative AI search is typically where research starts. It also found that AI answer engines often deliver incomplete or unreliable information, which creates mistrust, and that buyers turn to colleagues, peers and experts to validate what they find.

6sense's 2025 Buyer Experience Report (vendor research, nearly 4,000 buyers) found that buyers made first contact about 61% of the way through their journey, earlier than the 69% the company reported the year before, and that buyers initiated 79% of those first contacts.

Together: buyers prefer to work alone, they start with AI, they do not fully trust what AI tells them, and when they reach out they are well informed and usually already leaning.

Consistency is now a trust test

When a buyer has read your website, an AI summary of your website, three review sites and a competitor's comparison page before the first call, every inconsistency becomes visible. Gartner's 2025 survey found that 69% of buyers reported inconsistencies between the information on a supplier's website and what its sellers told them.

Inconsistency costs more than it used to for two reasons. Buyers are risk-averse (Forrester's description) and treat a contradiction as a sign of what the supplier will be like after the contract is signed. And AI tools amplify whatever is published, outdated pages included: a pricing page you changed last year can still shape an AI answer today.

The practical response is unglamorous: one written set of approved claims, kept current, that marketing, sellers and any automated outreach all draw from. Product capabilities, pricing principles, integration lists, security posture, named customer results that you have permission to share. When a seller wants to say something that is not on the list, the right move is to add it to the list, with its evidence, so the next buyer hears the same thing.

What the seller is for now

If information is self-served, the seller's value moves to three places.

Fit and risk. Buyers told Gartner they still want seller input on whether a product fits their specific situation. That is diagnostic work: understanding their environment well enough to say honestly where you are strong, where you are weak, and what would have to be true for it to work.

Validation. Forrester's buyers seek proof from people they trust. A seller who can connect a buyer with a peer customer, share a detailed implementation plan, or walk through the reasoning behind a claim provides the validation an AI summary cannot.

Consensus. As Lessons 2.1 and 2.2 argued, a buying group needs a shared definition of the problem. The seller who supplies it, and arms the champion to carry it, is doing work no self-service channel does.

What the seller is not for is repeating the website. A first conversation that opens with a capability overview tells a well-researched buyer that you did not expect them to have done their homework.

When the buyer reaches out

If buyers initiate most first contacts, and the winner is usually already on their shortlist, then the moment a buyer writes to you is precious. Respond quickly, and respond specifically: acknowledge what they asked, show you looked at who they are, and propose a concrete next step. In 2011, James Oldroyd, Kristina McElheran and David Elkington reported in Harvard Business Review that most companies were not responding to online leads nearly fast enough. The specific figures from that study are widely repeated without sources, so we will not quote them, but the principle has only become more pressing as buyers arrive later and more decided.

Kavanah helps on both halves. Synced mail lands in /inbox, and in the Deal Room the Conversations panel's 'We owe a reply' filter lists every deal contact whose last email is waiting on you, oldest first, so a buyer who wrote nine days ago is not buried under one who wrote this morning. For the consistency half, keep your approved claims in a Knowledge document; the agent can search the workspace's Knowledge when it drafts, and a draft built from the approved list is far less likely to contradict your website than one built from a seller's memory.