Breeze Customer Agent

Breeze Customer Agent is the support agent in HubSpot's Breeze AI family. Assigned to a support channel, it reads the customer's question, answers from the knowledge sources you gave it, asks a follow-up when the question is unclear, and reassigns the conversation to a human when its confidence is low or a handoff rule fires. Every answer is meant to be backed by a source it can cite, such as a knowledge base article. It is one of three agents most HubSpot teams start with, alongside the Prospecting Agent for sales and the Data Agent for answers about records; our comparison of Customer vs Prospecting vs Data covers which to start with.

What it needs, from HubSpot's set up the customer agent article (last updated by HubSpot 16 September 2026, read 29 September 2026): a Professional or Enterprise subscription on Marketing, Sales, Service, Data, Content or Revenue Hub, or Smart CRM; HubSpot Credits to deploy it to channels; the Customer agent editor permission and an assigned seat for whoever builds it; at least one live chat, WhatsApp or Facebook channel connected to the inbox or help desk; and the generative AI, CRM data, conversation data and files data switches on in AI settings. First-time accounts can opt in to 14 days of free access, during which it uses no credits. Once created, a customer agent can be paused but not deleted, so name it for the brand or channel it serves.

What it costs is a resolution rate, not a seat. HubSpot's customer agent overview (updated 16 September 2026) counts a conversation as resolved when the agent replied with a content source or performed an action and the visitor did not ask for a human within 72 hours, or when the agent marks a lead qualified, partially qualified or not qualified. The published rate is 50 credits per resolution on a text channel, as listed on INSIDEA's agent activation page and transcribed from HubSpot's product catalog; at $0.010 a credit, that is $0.50. Reply recommendations, where the agent drafts an answer in the help desk for a rep to send, use no credits at all, which makes them the lowest-risk way to test the agent's answers on real tickets. The AI credits breakdown has the arithmetic for other volumes.

The common mistake is switching it on before the knowledge is ready. An agent that crawls a whole domain answers support questions in marketing copy, and one with no written handoff rules keeps customers talking to software when they want a person. Match specific help-centre paths, write the handoff as a rule, and test the last quarter's most common questions in preview, which costs nothing, before deploying to one channel. INSIDEA's Customer Agent activation does exactly this on a portal that already has the agent, as a fixed fee from $250 on the ten-day Standard track, with a credit forecast before go-live and the actual usage reviewed after the first 24 hours. The Customer Agent playbook covers the full setup.