Yes. AI agents can qualify and route leads using HubSpot CRM data, and in 2026 you can do it with HubSpot's own agents in Agent Hub, with AI actions inside workflows, or with a third-party agent connected to your account. The right choice depends on where your leads come from (chat, forms or outbound lists), which HubSpot plan you have, and how many HubSpot Credits you want to spend.
Checked against the HubSpot Knowledge Base, HubSpot developer docs and the HubSpot Product & Services Catalog on October 20, 2026.
This guide covers the native agents that matter for leads, the workflow actions that let you build your own qualification logic, how routing works, what each option costs in credits, and a step-by-step build you can copy.
The four ways to qualify and route leads with AI in HubSpot
| Option | Best for | What it does with leads | Plan needed |
|---|---|---|---|
| Customer agent (lead qualification, beta) | Inbound chat, Facebook or WhatsApp conversations | Qualifies visitors against your criteria, sorts them into Qualified, Partially qualified or Not qualified, then books a meeting or updates lifecycle stage | Professional or Enterprise |
| Prospecting agent | Outbound and target accounts | Watches target companies for intent signals, suggests contacts and writes personalized outreach | Starter, Professional or Enterprise (Smart CRM needs Pro or Enterprise) |
| Workflows with AI actions | Form fills and any record-based trigger | Data Agent actions or the Run agent action return an answer you can branch on, then Rotate record to owner assigns the rep | Professional or Enterprise |
| Custom code or third-party agents | Logic HubSpot's tools do not cover | Your own code or an external agent reads and writes HubSpot records | Data Hub Pro or Enterprise for custom code; app permissions for external agents |
You do not have to pick one. You can combine them, for example the customer agent for chat, a workflow for form fills, and lead scoring underneath both.
Current names: what HubSpot calls its AI tools now
HubSpot has renamed several AI tools, and older guides still use the old names. As of October 2026, HubSpot's own pages use these names:
- Breeze Assistant is the conversational assistant. You open it from the Breeze Assistant button in the top navigation bar. It can generate content, summarize records and help you prepare for meetings. An admin must first turn on Give users access to Breeze Assistant in AI settings. Older names such as Breeze Copilot and ChatSpot no longer appear in HubSpot's help article for it.
- Agent Hub is where agents and agentic workflows live. Go to More > Agents > Agent Hub (or Agents > Agent Hub if you do not see More). It has Intro, Agents and Context tabs.
- Agent builder is the tool for creating custom agents, opened from Create > Agent in Agent Hub. HubSpot's current help articles call it the agent builder.
- HubSpot-built agents listed on HubSpot's AI product page include the Content, Prospecting, Customer, Campaign, Nurture, Data and Revenue agents.
The Context tab is worth setting up before any lead work. It holds the business information agents use, such as messaging, tone, brand and ideal customer profiles, and knowledge vaults hold more detailed material like product details.
The native agents that matter for lead qualification
Customer agent: qualify leads inside conversations
The customer agent responds to support conversations and resolves customer requests. Its lead qualification feature is in beta, and a Super Admin can opt the account in.
How it works, according to HubSpot's setup article:
- Go to More > Service > Customer Agent, then open Train > Actions.
- In the Lead qualification section, click Edit.
- In Trigger instructions, describe when the agent should treat a visitor as a lead.
- Add qualification criteria as property filters (property, operator, value). You can create a new property from the same screen.
- Set Lead routing rules for each category. The documented options are booking a meeting with a selected meeting link, or updating the lifecycle stage.
The agent sorts each lead into Qualified, Partially qualified or Not qualified and stores the result in the Customer Agent Lead Status property. That property is useful because a workflow can pick it up and handle owner assignment, which the lead routing rules themselves do not document.
Plan and channel needs: Professional or Enterprise in Marketing, Sales, Service, Data, Content Hub, Smart CRM or Revenue Hub, plus at least one Facebook, WhatsApp or live chat channel connected to the inbox or help desk.
Credits: setting up lead qualification does not use credits. The agent charges per resolution, not per message, and marking a lead as Qualified, Partially qualified or Not qualified counts as a resolution. HubSpot's catalog lists 50 credits per resolved text conversation.
Prospecting agent: research and outreach for target accounts
The prospecting agent is for outbound. You find it at More > Sales > Prospecting Agent. You set it up by building and publishing a "play." The agent:
- Monitors companies in your target segments for intent signals such as funding news, product launches or research on topics you choose.
- Suggests relevant contacts at matching companies. You can also connect one external data source (Apollo, Surfe, ZoomInfo or Seamless).
- Writes personalized emails, using steps it decides or a sequence you configure.
- Sends from the contact owner, the company owner or one user, and you can require review of every email before it sends.
Credits: each lead that receives generated outreach uses credits, whether the agent suggested the lead or you enrolled it. The catalog lists 100 credits per lead outreach recommendation. HubSpot also offers 14 days of free access for eligible Professional and Enterprise accounts in the related beta, starting when you publish your first play.
Data agent: fill the fields your qualification depends on
Qualification is only as good as the data behind it. The data agent, at More > Data Management > Data Agent, creates smart properties that hold AI-generated answers. Each smart property pulls from web research, the company website, property data or call transcripts. You can fill smart properties in bulk, on a schedule for a segment, or from a workflow.
Credits: the catalog lists 10 credits per prompt response for one record.
Build your own qualification logic with workflows
If your leads arrive through forms, a workflow is usually the cleanest route. HubSpot documents three Data Agent workflow actions, plus a Run agent action in the AI section of the workflow editor.
| Workflow action | What it does | Useful for |
|---|---|---|
| Data Agent: Custom prompt | Analyzes, summarizes or categorizes the enrolled record's data from a prompt you write (up to 3,000 characters). It has no internet access. | Classifying a lead as high fit, nurture or unqualified |
| Data Agent: Research | Answers a prompt from the record's properties or its five most recent call transcripts | Checking what was said on discovery calls |
| Data Agent: Fill Smart Property | Writes a value to a smart property | Enriching industry, use case or company type |
| Run agent | Runs an agent from your account and returns a text response or structured outputs (text, number, boolean, date, enumeration and more) | Company research before routing, or category outputs you can branch on |
To branch on an AI answer, add a Branch action, choose One property or action output, and select the action's response. You can also write the response to a property with an Edit record action.
Requirements: these actions need a Professional or Enterprise subscription and HubSpot Credits. The Run agent action has a limit of 500 executions per day. If the account runs out of credits, the Data Agent: Custom prompt action fails and returns null values, so build a fallback branch.
Agent Hub also has its own workflow builder. From the Workflows tab in Agent Hub, you can click the + Add icon, open the AI tab and choose Add an agentic step to run an existing agent or create a new one from a prompt. Agentic steps must connect to the main sequence of triggers and actions, or they will not run.
Routing: assign the owner
Once a lead has a category, use the Rotate record to owner action to assign it. HubSpot's routing article says it needs Sales Hub or Service Hub Professional or Enterprise, and owners must be active users with paid Sales Hub or Service Hub Professional or Enterprise seats. You can rotate across specific users or a team. For lead records (beta), you can choose load balanced, round robin or random distribution.
A practical pattern is to branch by territory, company size or industry first, then add a rotate action inside each branch.
Custom code for anything else
If you need logic HubSpot's AI actions do not cover, such as calling your own model or an outside scoring API, the Custom code workflow action needs Data Hub Professional or Enterprise. It supports JavaScript (Node.js), with Python in beta. Each run must finish within 20 seconds and use no more than 128 MB of memory, and API keys belong in secrets, which the code reads as environment variables.
Step by step: an AI lead qualification and routing workflow
This is a hypothetical example for a B2B company with a "Talk to sales" form. Adjust the criteria to your own ideal customer profile.
- Write down the rules first. Define high fit, nurture and unqualified in plain language (for example, company size, industry, role and stated timeline). Put the same profile in the Agent Hub Context tab.
- Make sure the inputs exist. Use the data agent to fill smart properties such as industry or company type for leads that arrive with thin data.
- Create the workflow. Go to Automation > Workflows, create a contact-based workflow, and set the trigger to the form submission.
- Classify with AI. Add Data Agent: Custom prompt or Run agent with structured outputs. Ask for one category from a fixed list (High fit, Nurture, Unqualified) and a one-line reason, and save both to properties with Edit record.
- Branch on the answer. Add a Branch on the action output. Add a fourth path for empty answers, so a credit shortfall or a blank response goes to a human instead of disappearing.
- Route. In the High fit path, use Rotate record to owner and create a task or notification for the rep. Send Nurture leads to a nurture sequence or list. Mark Unqualified leads, but do not delete them.
- Test before you spend. Test custom agents in the agent builder (testing does not use credits), and run the workflow on a small set of sample records first.
- Cap usage. In Agent Hub, open the agent, click the three dots and choose Edit run limit to set a monthly run limit.
- Review weekly. Compare the AI categories with what reps see on calls, and tighten the prompt or criteria where they disagree.
Lead scoring still matters
AI qualification works best on top of a lead score, not instead of one. HubSpot's lead scoring tool (Marketing > Lead Scoring) needs Marketing Hub or Sales Hub Professional or Enterprise. Contact scores need Marketing Hub, company scores work with either hub, and deal scores (combined only) need Sales Hub. Fit scores rate records by property values, engagement scores rate them by events, and combined scores use both. Combined scores add a label from A (high fit) to C (low fit) and 1 (high engagement) to 3 (low engagement), so a low-fit but highly engaged contact is C1.
Building contact scores with AI needs Marketing Hub Enterprise and a sample of at least 50 contacts (25 converted and 25 not converted).
What it costs in HubSpot Credits
HubSpot's Product & Services Catalog lists these included monthly credits and costs (checked October 20, 2026):
| Plan | Starter | Professional | Enterprise |
|---|---|---|---|
| Smart CRM, Marketing, Sales, Service, Content or Revenue Hub | 500 | 3,000 | 5,000 |
| Data Hub or Customer Platform | 500 | 5,000 | 10,000 |
| Action | Credits |
|---|---|
| Customer agent, resolved text conversation | 50 |
| Prospecting agent, lead outreach recommendation | 100 |
| Data agent, one prompt response for one record | 10 |
| One AI action in a workflow | 10 |
| Custom agents (beta) | 1 per action unit |
Included credits reset monthly and do not roll over. If you have several products, they do not stack; you get the highest allotment. As a rough example, 3,000 included credits would cover about 300 single AI workflow actions in a month, or about 60 resolved customer agent conversations, before other usage.
Third-party agents and HubSpot
You can also run lead qualification in an outside agent platform. HubSpot's remote MCP server at `https://mcp.hubspot.com` lets external AI tools read, write and update HubSpot records as a user would. It uses OAuth with PKCE, and HubSpot says all actions respect the connected user's existing HubSpot permissions. In the other direction, HubSpot's agent builder can give your own agents tools from external systems connected through MCP.
Some tools that are not lead routing tools also connect to HubSpot. For example, Profound, an AI visibility platform, said in its February 24, 2026 funding announcement that its agents integrate with HubSpot. That announcement describes agents built on AI visibility data, not lead qualification, so check any vendor's documentation for what its HubSpot connection can read and write before you rely on it for routing.
Before connecting any third-party agent, confirm three things: which HubSpot objects it can change, whether a person reviews its actions, and where it logs what it did.
Common mistakes to avoid
- Letting AI pick from open-ended answers. Ask for a fixed list of categories so branches work every time.
- No fallback path. Empty AI responses and credit shortfalls happen. Route them to a person.
- Skipping the data step. An agent cannot judge fit from a blank company size field.
- Assigning to users without paid seats. Rotate record to owner excludes users without a paid Sales Hub or Service Hub Professional or Enterprise seat.
- No spending cap. Set monthly run limits per agent and review estimated credits in run history.
If you want help designing this inside your own portal, INSIDEA is an Elite HubSpot Partner and works on HubSpot setup and RevOps as the growth operating system for modern businesses. You can see how we approach it on our HubSpot services page and RevOps page.
How INSIDEA works
