By
After HubSpot's fall release there is suddenly an agent for almost everything. One that prospects, one that moves deals forward, one that answers customers, one that writes content, one that tailors every nurture email and one that chases unpaid invoices. The question many people are asking is no longer whether to use AI agents in HubSpot, but where to start.
In short: HubSpot's AI agents live in Agent Hub and cover prospecting, deals, campaigns, content, nurturing, customer service, invoice follow up and data quality. Start with one agent, one clearly defined use case and human approval on everything, once Context Home and your CRM data are in place.
Here I go through the most important agents presented at UNBOUND 2026, what they actually do and how to introduce them without automating the chaos that already exists in your CRM.
Agent Hub: the control room
All agents are now gathered in Agent Hub, where you see which are active, what they do and what results they deliver. There you also find Agent Builder (formerly Breeze Studio) for building your own agents. What is new is that agents can be triggered from more places, such as Slack, a Google Sheet, a webhook or a schedule, and that older workflows can be migrated to agent based ones with one click.
The agents at a glance
For sales
Prospecting Agent monitors more than 40 buying signals, researches target companies, assembles a buying group and writes personal outreach in your tone of voice. It learns from the edits reps make.
Deal Progression uses emails and meeting transcripts to suggest next steps, draft replies, create quotes and update CRM fields. You choose what can happen automatically and what requires approval. Requires Professional or higher and is in public beta.
For marketing
Campaign Agent turns a goal or an insight into a coherent multi channel campaign plan. One example shown was feeding in a sales call transcript and getting a brief, assets and automation out.
Content Agent researches brand, audience and competitors and creates emails, ads, landing pages, blog posts and LinkedIn posts.
Nurture Agent rewrites the subject line, body and CTA for each recipient at the moment the email is sent, based on CRM data, engagement and research about the company. It sits on top of your existing workflows.
For customer service and finance
Customer Agent got 17 updates, including better routing, handover to a human in the middle of a conversation, tone of voice per segment and the ability to create knowledge base articles from resolved tickets.
Revenue Agent is new and follows up unpaid invoices, sends reminders and answers questions. If a late payment turns out to be caused by a support ticket or a dispute, it adapts its behavior.
For your data
Data Agent enriches and maintains CRM data. It is less glamorous than the others, but it may well be the most important one.
Five pitfalls to avoid
1. Starting with the agent instead of the process
An agent follows the rules it is given. If your ICP, your qualification criteria, the handover between marketing and sales and the logic of your nurturing are not clearly defined, the agent will do something, but hardly what you want. Ask yourselves first: could a new employee do this job based on what we have documented?
2. Automating bad data
Agents inherit every problem that already exists in the CRM and apply them faster. A prospecting agent working on a list full of duplicates and wrong job titles sends the wrong message to the wrong person, at high speed.
3. Skipping Context Home
Every agent reads from Context Home. If the context is thin or outdated, the results will be generic. Review and complete it before you activate anything.
4. Letting everything loose at once
Start with approval on everything. Let the agent suggest and the human decide. When you see that the suggestions hold high quality over time, you can gradually release the automation for simpler tasks.
5. Forgetting GDPR and transparency
Agents that make decisions about individuals, such as which message someone should receive or how a payment reminder should be phrased, touch on the rules for automated decision making. Document what each agent does, what data it uses and where a human comes in. It is a good idea to check with your data protection officer before going live.
How I would start
- Weeks 1 to 2: Review and complete Context Home. Run a data quality check of contacts, companies and deals.
- Weeks 3 to 4: Choose one clearly defined use case with measurable value. For many B2B companies, Deal Progression or Prospecting Agent against a well defined list of target accounts is a good first step.
- Weeks 5 to 8: Test the agent on a fixed number of real cases, with approval on everything. Measure time saved and the quality of the suggestions.
- After that: Evaluate, adjust context and rules, and take on the next agent.
Also keep in mind that several of the features consume HubSpot credits. Keep track of usage before you scale up.
Agent or workflow: when should you use which?
A common question is whether agents replace workflows. They do not. They solve different types of problems.
A workflow follows rules you have decided in advance. If a contact fills in a form, set the lifecycle stage and send a certain email. It is predictable, cheap and easy to troubleshoot. Use workflows when you know exactly what should happen.
An agent gets a goal and some input and decides for itself how to solve the task, for example which message suits a particular recipient or which next step makes sense in a deal. That gives more flexibility, but also more variation in the result. Use agents when the task requires interpretation, research or text adapted to the situation.
The rule of thumb we give our clients: anything that can be described as "if this, do that" belongs in a workflow. Anything that requires judgment is a candidate for an agent, with a human approving until the quality is proven.
Frequently asked questions about AI agents in HubSpot
Do the agents replace sales reps and marketers?
No. They take over repetitive steps such as research, updating fields and first drafts. People provide judgment, relationships and decisions. Used correctly, they free up time for what actually moves deals.
Can we build our own agents?
Yes, through Agent Builder. With the new triggers and the ability to connect third party apps, you can build agents for your specific processes. Still, it is a good idea to start with one of HubSpot's ready made agents to learn how it works.
What happens to our existing workflows?
They keep working as usual. There is a migration to agent based workflows, but not every workflow benefits from being moved. Pick the ones where flexibility and personalization actually make a difference.
What is an AI agent in HubSpot?
An AI agent in HubSpot performs a task on its own based on a goal, your CRM data and the context in Context Home. Unlike a regular chatbot it can act, for example research a company, write outreach or update CRM fields.
How much do the AI agents in HubSpot cost?
Several of the agents consume HubSpot credits, and which ones are included depends on your hubs and tiers. Review your credit usage in the portal before you scale up.
Which AI agent should you start with?
Start where you have a clear process and measurable value. For many B2B companies, Deal Progression or Prospecting Agent against a defined list of target accounts is a good first step. If you have large volumes in customer service, Customer Agent can deliver the fastest impact.
Are HubSpot's AI agents GDPR compliant?
It depends on how you use them. Document what data each agent uses, what decisions it makes and where a human reviews. Agents that make decisions about individuals should be checked with your data protection officer.
What is Agent Hub in HubSpot?
Agent Hub is the place in HubSpot where all AI agents are gathered. There you activate, monitor and build agents, and it is also where Context Home lives. Agent Hub replaces what used to be called Breeze Agents.
Summary
The AI agents in HubSpot are real now, and they can save a lot of time. But they amplify what already exists. Clear processes, clean data and well developed context get better. Unclear processes and messy data get worse, just faster.
Want help choosing the right first agent and setting it up in a way that lasts? Get in touch with LeadLabs and we will go through your portal together.
