HubSpot’s agents have charged by the finished job since April. What changed at Unbound, held in Boston from 16 to 18 September, is that HubSpot now wants that logic to shape how every customer pays. At its analyst day on 17 September, the company described a hybrid model: seats cover the people using the platform, and credits cover the work its AI agents do.

The shift matters because the credit line is the part of the bill that grows with use. HubSpot launched Campaign, Content, Nurture and Revenue agents at the same event, and told investors that monthly agentic actions have risen 3.5 times this year. For marketing teams, the question is moving from how many licences they hold to how much work they let the agents do.

What did HubSpot actually change at Unbound?

HubSpot did not introduce per-task pricing at Unbound. That arrived on 14 April, when the Customer Agent moved from $1 (about £0.75) per conversation to $0.50 (about £0.37) per resolved conversation, and the Prospecting Agent moved from a monthly charge per enrolled contact to $1 per lead recommended for outreach. Chief customer officer Jon Dick said at the time: “You pay when it works, full stop.”

A State of AI report published on 16 September presented the per-task billing as news from the conference. An independent recap from HubSpot partner blog Hacking Demand was plainer: “The Spotlight release did not announce pricing changes.”

The new commercial detail came at the analyst day. HubSpot said it is piloting the seats-and-credits model with new customers in the Nordics and Benelux, with plans to extend it to other EMEA regions, according to reports from MarketBeat and Martech Notes. HubSpot has not given a date for the UK. Investors were unimpressed: the shares fell 4.3% on 17 September, according to Martech Notes, as analysts looked for clearer signs of near-term revenue growth.

How much does HubSpot’s AI cost per task?

HubSpot prices credits at $0.01 each, sold in packs of 1,000 for $10 (about £7.50). Its Marketing Hub pricing page lists 50 credits per resolved Customer Agent conversation, and HubSpot’s April announcement set the Prospecting Agent at 100 credits per recommended lead.

The included allowances are modest. Marketing Hub Professional, at $800 (about £600) a month on an annual contract, comes with 3,000 credits. That covers 60 resolved conversations a month and nothing else. Included credits expire at the end of each month, and overage defaults to pay-as-you-go, so heavier use flows straight into the invoice.

The Nordic pilot suggests where the maths is heading. According to Martech Notes, which reviewed HubSpot’s Nordic product catalogue supplement, the pilot launched on 22 June with seats from €20 a month and much larger monthly credit allowances: 5,000 on Starter, 10,000 on Professional and 15,000 on Enterprise. It also prices Content Agent at 1,000 credits per asset, roughly £7.50 for each piece the agent generates.

The outcome model can cut costs. HubSpot says its Customer Agent resolves 65% of conversations. On that figure, 100 conversations that cost $100 under the old pricing would cost $32.50 under the new one. The risk sits elsewhere: a single credit pool shared across marketing, sales and service, drawn down by a growing number of agents. Hacking Demand’s advice to HubSpot customers was blunt: “Agent sprawl burns credits the same way tool sprawl burned budget.”

HubSpot does provide the brakes. Admins can set an account-wide cap, limits for individual agents or features, and usage alerts, and can pause any credit-based feature, according to HubSpot’s credits page.

Does HubSpot’s AI model lock marketing teams in?

HubSpot’s critics argue that it does. The State of AI piece summarised a LinkedIn post, which it did not name, arguing that a conference called Unbound spent its keynote keeping AI inside HubSpot. On that view, customers build agents in HubSpot, feed them HubSpot data, manage them in HubSpot and pay HubSpot each time one acts. The Helm could not verify the original post.

HubSpot’s own announcements point the other way, at least for data. On 16 September it said weekly active users of its ChatGPT connector have grown 250% since the connector launched in July 2025. HubSpot did not give a base figure, so the percentage says little about absolute scale. It also became the first CRM to integrate with ChatGPT Ads, letting teams build, run and measure ChatGPT campaigns alongside their other channels. And it launched an AI Growth Bundle with OpenAI: HubSpot Starter and credits at up to 65% off for a year, a buy-one-get-one-free offer on ChatGPT Business seats, and a $750 (about £560) match on ChatGPT Ads spend for new ad accounts opened by 30 September.

The two accounts are less contradictory than they look. HubSpot is opening its data to outside AI tools while keeping the billing for agent work inside its own credit system. Chief executive Yamini Rangan framed the analyst day around what she called the “outcomes era”, arguing that business context is “what turns that intelligence into measurable customer outcomes”. The context, and the meter, stay with HubSpot.

HubSpot customers have pushed back on terms before. In July, the company withdrew new data-sharing terms four days after a customer backlash.

HubSpot has said the seats-and-credits model will reach more EMEA regions but has not named them, published UK pricing or said whether existing customers will be moved onto it.