What changed
On April 14, 2026, HubSpot moved two of its Breeze AI agents from subscription-style pricing to outcome-based billing. Breeze Prospecting Agent, which researches accounts, identifies buying-committee contacts and drafts outbound emails inside Sales Hub, had been billed as a recurring monthly charge per enrolled contact. It now costs $1 for each lead the agent recommends for outreach. Breeze Customer Agent moved from $1.00 per conversation to $0.50 per resolved conversation (HubSpot, 2026).
Both agents are available on Sales Hub Starter, Professional and Enterprise, paid through HubSpot Credits at $10 per 1,000 credits, with a 28-day free trial (HubSpot, 2026). The mechanics are simple: no qualified lead, no charge for that lead. No resolved conversation, no charge for that conversation.
HubSpot has not published independently audited results, but the company reports that across more than 8,000 customers, Customer Agent resolves 65% of conversations and cuts resolution time by 39%, while Prospecting Agent has seen a 57% quarter-over-quarter rise in activation and is associated with a 10% lift in close rates (HubSpot, as reported by CX Today, 2026). These are self-reported vendor figures, not third-party verified, and should be read as directional.
Why vendors are moving off seats
HubSpot is not first. Zendesk called its shift "the industry's first outcome-based pricing" for customer service AI when it launched resolution-based billing in August 2024, charging $1.50 per automated resolution (Zendesk, 2024). Salesforce launched Agentforce at Dreamforce in September 2024 with a flat $2 per conversation, then replaced that model in May 2025 with Flex Credits: $500 per 100,000 credits, with a typical agent action consuming 20 credits, or $0.10 (Salesforce, 2025). Intercom's Fin AI Agent charges $0.99 for a resolution, handoff or disqualification, and $9.99 when Fin qualifies a prospect and routes it to a rep (Fin, 2026).
The logic is the same across all four: an AI agent doing a junior rep's or support rep's job is priced against the work it produces, not against a seat someone has to log into. For a sales organization, that reframes AI spend as a cost of goods sold on pipeline rather than a software line item, which changes who signs off on it and how it gets forecast.
How the models differ
The four vendors above land on different units of work, and the unit matters more than the headline rate. HubSpot bills per qualified lead surfaced for outreach, a definition set by HubSpot's own AI against whatever ICP criteria the customer configures. Zendesk and Intercom bill per resolved support interaction, a comparatively easy outcome to define. Salesforce's Flex Credits bill per discrete backend action, which is granular but harder for a buyer to predict in advance since a single customer interaction can trigger several billable actions.
For an SDR or AE team, the practical difference is this: a per-qualified-lead model ties cost directly to pipeline volume, so spend rises and falls with how much the agent prospects, not with headcount. A per-action or per-conversation model is closer to traditional usage-based software billing, harder to tie to a specific sales outcome.
What to check before buying
Get the vendor's exact definition of the billable outcome in writing before signing. HubSpot's "qualified lead" is whatever HubSpot's AI decides matches configured buying signals, not a rep's own judgment call, and a loosely configured ICP will inflate both lead volume and the invoice. Ask what happens to already-drafted outreach when a credit balance runs out mid-quarter, since a stalled prospecting agent in the last two weeks of a quarter is a different problem than a stalled dashboard. Ask whether pipeline sourced by the agent counts toward rep quota and commission the same way rep-sourced pipeline does, since that decision affects comp plans, not just tooling.
What commonly goes wrong
The most common failure mode is treating the outcome definition as fixed when it is configurable. A prospecting agent tuned to a wide ICP will generate more billable "qualified" leads with lower actual fit, and the vendor has no financial incentive to tighten that bar. A second failure is budget ownership: outcome-based AI spend often lands with sales ops or RevOps rather than IT, and if nobody owns the monthly credit forecast, teams find out they are over budget only when prospecting stops. A third is comparing sticker rates across vendors without normalizing for what counts as one unit of work, since $1 per HubSpot lead and $2 per Salesforce conversation are not measuring the same thing.