HubSpot Customer Agent answers customer questions using information your business provides. On supported customer channels, it can respond from that information, ask for clarification or pass the conversation to your team. It is a way to help customers get useful answers while giving people a clear route to human help. HubSpot’s setup guide describes the feature.
If you are still getting familiar with the terminology, our HubSpot agents overview explains how agents, Breeze Assistant and workflows fit together.
For a business owner, the first decision is specific: which repeated question would you trust an agent to help answer, and who takes over when that answer is not enough?
Imagine a customer asking what to prepare before an onboarding appointment. Your current guide explains the documents to bring, the expected duration and how to request a different time. The agent could use that guide to explain the preparation steps.
Now imagine the customer says, “My situation is different. Can someone help me?” A useful experience needs more than another automated reply. It needs an agreed destination for that request and an accurate expectation about when someone will respond.
This is an illustrative scenario, not a claimed Foundry 11 customer result. Its value is in making the job easy to assess: did the customer receive a correct answer or reach the right person?
Choose one group of common questions with answers that are easy to check. Appointment preparation, standard service information or a documented support process may be suitable starting points.
Before adding sources, review them as though you were sending them directly to a customer. Check dates, exclusions, support hours and conflicting instructions. Give each source an owner. When a policy changes, that person should know which material needs updating.
Our recommended first pilot is deliberately narrow. Write down the questions it covers and the situations that should go to your team. Review sample answers against the original sources, including qualifications that could change a customer’s decision.
Customer Agent can also access or update selected CRM properties when that access is configured. That is a separate capability to assess; supplying support content does not mean every customer record should be available to the agent. See HubSpot’s CRM access guide.
For a first pilot, decide whether answers alone are enough. If an update is needed, name the permitted field, the reason for the change and the person responsible for correcting errors. Check the underlying records before expanding access. Our CRM cleanup checklist can help frame that work.
HubSpot provides configurable handoff rules, routing and messages. Its handoff guide explains the available arrangements, including what happens when someone needs human help.
Our recommendation is to test the receiving team’s experience as carefully as the customer’s. Confirm who sees the conversation, whether they have enough context and who owns it until completion. Check the out-of-hours message against actual staffing. An agent being available after hours does not mean a person is waiting to respond immediately.
Use a short acceptance checklist:
| Test | A useful result |
|---|---|
| A familiar question | The answer matches the current source, including exclusions. |
| A question without a documented answer | The agreed fallback is used rather than an invented answer. |
| A direct request for a person | The request reaches the intended team or follow-up queue. |
| A request outside support hours | The response-time expectation matches your service. |
| A permitted record update, if enabled | Only the intended information changes. |
Current setup documentation lists eligible Professional and Enterprise subscriptions, an assigned seat and relevant permissions. Deployment uses HubSpot Credits; check your account’s current terms before budgeting. Our HubSpot AI costs checklist covers the commercial questions.
For the pilot, record correct answers, useful handoffs, missed follow-ups and the time your team spends checking or correcting work. Agree who can pause it if repeated errors appear. Automated conversation volume alone will not tell you whether the customer experience improved.
Start with one repeated question, a reliable answer and a named person who owns exceptions. Widen the scope only after reviewing real conversations and the work left for your team.
Have a customer question you answer every week? Talk it through with Foundry 11. We can help assess the sources, account requirements and human follow-up before recommending a rollout.
It can interpret a customer’s question and respond using supplied information. You still need to define its scope and test how it handles uncertainty.
Plan for people to own exceptions, source updates and follow-up. Whether the pilot helps depends on the quality of the whole service process.
Check whether customers get accurate answers and completed follow-up, alongside credit usage and the effort required to review the work.