The 'Contact Sales' Button Is Becoming a Revenue Leak

A buyer asks an AI assistant to compare three customer-support platforms and report back with pricing, features, and a recommendation. The agent clicks through each vendor's pricing page and hits the same wall: "Contact Sales." It does not fill out a form. It goes to G2, to analyst write-ups, to comparison blogs, and serves those back to the human as the vendor's price. The vendor, who never saw the buyer, has lost control of the pricing conversation.

That is the new shape of a B2B software sale, according to a Forrester analysis by analysts Lisa Singer and Beth Caplow. The pricing page that used to be a sales funnel, opaque on purpose and designed to push a human prospect into a call, is now a routing failure for the agent that already represents that human. When the agent cannot read the price, it does not stall. It finds someone who can.

AI buying agents are software, not people. They don't enter forms, book demos, or submit credit cards. A "Contact Sales" button is a form gate, and a form gate is a dead end for an agent. When the agent hits one, it falls back to the next best source: third-party review sites, comparison blogs, and analyst write-ups. Those pages are designed to be machine-readable, and they are also written by someone with a different incentive than the vendor's.

The number behind the warning comes from Siteline, a vendor in the pricing-page infrastructure space, which benchmarked AI agents attempting to retrieve pricing and packaging from B2B software vendor sites. In roughly 30% of runs, the agent hit at least one search or retrieval error. The Siteline data is a single benchmark from a vendor with commercial interest in the fix, and Forrester's framing leans on it, so the headline number is best read as a directional signal rather than an industry census. But the behavior it describes is consistent with what the agents actually do: when the primary source is unreadable, they move to a readable one and report back as if the second source were the first.

That is the second-order consequence the Forrester piece names. B2B pricing has long been deliberately opaque, a strategy for sorting serious buyers from tire-kickers and protecting sales engagement from price-shoppers. The strategy assumed a human at the other end of the funnel. When the human is replaced by an agent, opacity does not sort. It routes. The agent does not internalize the vendor's "contact us for enterprise pricing" framing. It internalizes whatever G2, a Gartner Peer Insights write-up, or a comparison blog says the enterprise pricing looks like, and serves that to the human buyer as the answer.

The "Contact Sales" button still earns its place on genuinely custom deals: configuration, volume, regulated-industry terms, the kind of pricing that depends on a conversation. The category that needs to come off the page is the one the agent can already infer from public signals, including list price, per-seat cost, and the standard tier structure. The agent is finding those numbers anyway, on someone else's site. The vendors who publish them in a structured, machine-readable format keep the narrative in-house. The vendors who leave the page as-is hand it to the comparison site that ranks highest in the agent's fallback search.

The pricing page used to be a sales document aimed at a human. It is now the canonical source an agent reads first, and the canonical source the agent quotes when it talks to the human buyer. If that source is empty, the agent quotes someone else. The next time a B2B buyer asks an AI assistant to compare three vendors, the vendor whose pricing is on its own page is the one whose number the buyer will see. The vendor whose pricing is still behind a form gate is the one the assistant will not name.