HubSpot AI for Sales: What Breeze and the Prospecting Agent Actually Do

HubSpot’s AI tools can help sales teams perform five recurring jobs: identify active accounts, research and prioritize prospects, conduct outreach, prepare for and follow up on conversations, and maintain useful CRM data.

These capabilities do not operate as one autonomous salesperson. HubSpot’s AI sales tools are collectively branded Breeze—a collection of agents, assistants, and embedded features designed to handle different parts of the sales process. Some execute defined work independently. Others summarize information, recommend an action, or help a salesperson complete a task faster.

The practical value depends on matching the right tool to the work the sales team needs to remove, and understanding what each tool requires to perform reliably.

Key Takeaways

• HubSpot Breeze includes assistants, agents, and embedded AI features that perform different sales jobs.
• Buyer Intent identifies active accounts, while HubSpot Prospecting Agent researches and contacts prospects and Data Agent turns research into structured CRM information.
• Breeze can reduce the time salespeople spend on research, outreach, meeting preparation, follow-up, and CRM maintenance.
• Each tool requires specific inputs, including target-market criteria, reliable records, selling instructions, connected accounts, permissions, and HubSpot Credits.

Which HubSpot AI Tool Handles Each Sales Job?

The simplest way to understand HubSpot AI is to connect each capability to the work a sales team needs to complete.

A connected process might begin when Buyer Intent surfaces an active account. Enrichment or Data Agent can fill missing context, and Prospecting Agent can research and contact the appropriate person. Once that person books a meeting, HubSpot can summarize the record, capture the conversation, recommend follow-up, and suggest CRM updates.

No single tool controls the entire process. The tools remain distinct, but the data and activity connecting them live inside HubSpot.

1. How Does HubSpot AI Identify Active Accounts?

Buyer Intent helps sales and marketing teams find companies showing activity that may justify further investigation.

For first-party visitor intent, HubSpot uses its tracking code to collect website activity and match visits to companies where possible. Teams can create criteria around behaviors such as visiting a pricing, product, or consultation page within a selected period.

Buyer Intent can also surface broader signals, including:

• Companies researching selected topics across the web
• Funding announcements
• Executive hires
• Layoffs
• Product launches
• Mergers and acquisitions
• Other company news

Teams define target markets using criteria such as industry, location, and employee count. They then determine which pages, research topics, events, timeframes, and activity levels should influence sales attention.

Companies can be reviewed manually, added to the CRM, organized into saved views, or enrolled in workflows.

HubSpot Credits may be required to access features such as segment filtering and company exclusions, but using those features does not consume credits. Adding a net-new company from Buyer Intent to the CRM is the credit-consuming action. HubSpot’s Buyer Intent documentation explains the current requirements.

Because customers, competitors, job candidates, and researchers can produce similar activity, Buyer Intent narrows the field without confirming purchase intent. The sales team still decides whether an account is relevant and what should happen next.

HubSpot also offers buying signals within its newer Prospecting Agent experience. Teams can create plays that monitor selected companies for real-world triggers, identify relevant contacts, and generate signal-based outreach. This experience requires beta enrollment and should not be treated as part of every customer’s standard Prospecting Agent setup. HubSpot documents the beta separately.

2. How Does HubSpot AI Research and Prioritize Prospects?

Data enrichment, Data Agent, and Prospecting Agent address different parts of prospect research.

Data Enrichment Fills Known Information

Enrichment—powered by Breeze Intelligence—adds or updates standard contact and company information from HubSpot’s commercial dataset. Depending on what is available, that may include:

• Industry
• Employee count
• Location
• Company description
• Revenue range
• Contact title
• Other firmographic or demographic information

This creates a more complete record without requiring sales representatives to research and enter every field manually.

Data Agent Answers Organization-Specific Questions

Data Agent researches, analyzes, transforms, and structures information according to questions defined by the organization.

A sales or RevOps team might ask:

• Has this company raised funding in the past 12 months?
• Does its website indicate that it serves enterprise customers?
• Is it hiring for a role related to our product?
• Does the company meet our complete ICP criteria?

The answer can become structured CRM data rather than remaining inside a private AI conversation. Teams can then use it in segmentation, scoring, workflows, reporting, or outreach.

Data Agent currently supports:

• Smart Properties, which populate CRM fields with AI-generated answers
• Smart Actions, which run Data Agent inside workflows
• Smart Columns, which add AI-generated research or analysis to Data Studio datasets

HubSpot Credits are consumed when Data Agent generates an answer for a record. Smart properties can draw on CRM information, websites, web research, files, activities, and transcripts when the appropriate settings are enabled. HubSpot’s Data Agent documentation and smart-property documentation explain the current requirements.

The practical distinction is straightforward:

Enrichment fills known data fields. Data Agent answers a question or creates a classification specific to your business.

Fields that affect routing, disqualification, or high-value outreach should still include a validation process. Outdated websites, incorrect associations, and ambiguous prompts can all produce unreliable classifications.

3. How Does HubSpot’s Prospecting Agent Conduct Outreach?

Prospecting Agent can research selected contacts, generate personalized emails, and execute an outreach strategy within guardrails established by the organization.

Build a Selling Profile or Play

The agent needs clear instructions about what it represents and how it should approach prospects.

Depending on the Prospecting Agent experience available in the portal, teams configure a selling profile or a more detailed play that defines:

• The product or service being sold
• The intended audience
• The value proposition
• The sender identity
• The desired tone
• The call to action
• Supporting links or documents
• The signals or conditions that should influence outreach

A company can create different profiles or plays for different products, audiences, and positioning strategies.

This configuration matters more than whether an email sounds friendly or formal. A profile that tries to cover too many audiences can produce polished outreach without establishing a convincing reason to contact the recipient.

Establish Guardrails

Teams can control:

• Whether emails require approval
• Whether the agent can send automatically
• Which days and hours it can send
• How much time passes between emails
• How many emails an enrollment can include
• Which contacts should be excluded
• Which custom instructions govern the outreach

Review mode is the better starting point for most teams.

In the prospecting systems Hypha is building, the first review queue serves as a configuration test as much as an editorial review. Weak emails often reveal that the wrong contacts entered the play, company associations are incomplete, or the selling profile covers too many audiences. Reviewing the first batch of contacts gives the team a controlled way to correct those inputs before increasing volume or autonomy.

Select or Source Prospects

Contacts can be enrolled manually, through rulesets, through workflows, or from other eligible areas of HubSpot. Prospecting Agent can also draw on engagements associated with the contact record from up to the past year—including form submissions, page views, calls, meetings, notes, and email opens—to inform its research and outreach.

The buying-signals beta expands this further, letting the agent monitor selected company segments for triggers, source matching contacts through a connected data provider, and apply different plays to different audiences.

One current Hypha implementation applies this to a professional skincare brand with a prospecting universe of more than 20,000 contacts across med spas, aesthetic practices, dermatology groups, and vocational schools, using ZoomInfo to supply and enrich eligible records. The rollout follows the same phased logic: a small review batch first, a controlled monthly volume next, and signal-based plays only after the data and instructions have proven reliable.

Conduct Scheduled or Adaptive Outreach

Prospecting Agent supports two broad approaches:

• Scheduled outreach sends emails according to a defined schedule.
• Adaptive outreach changes the timing or content based on prospect behavior and engagement.

The agent can research the recipient, generate emails, and continue outreach within the established limits. When a recipient replies or books a meeting, the relationship returns to a salesperson.

Monitor Credits and Results

Prospecting Agent consumes HubSpot Credits per lead when it generates outreach—currently 100 credits, or $1, whether the lead was manually enrolled or surfaced through the buying-signals beta.

Credit rates can change. Teams should consult HubSpot’s current Product and Services Catalog and review the credit settings in their own portal before estimating usage at scale.

Credits reset monthly and do not roll over. If an account exhausts its included credits without additional capacity, credit-based features pause until the next reset or until more credits are purchased. Administrators can also set usage limits by tool. HubSpot’s credit documentation explains the current billing options.

That makes targeting a cost-control issue as well as a sales-performance issue. Enrolling the wrong contacts produces weak outreach and consumes credits on work the team did not need.

For a deeper look at how data, intent, qualification, outreach, and human review connect, read The Outbound Prospecting System for HubSpot.

4. How Does HubSpot AI Support Sales Conversations?

HubSpot’s AI capabilities continue after a prospect responds or schedules a meeting.

Breeze Assistant Summarizes CRM Context

Breeze Assistant can summarize contact, company, deal, and ticket records using information available to the user.

Before a meeting, a salesperson can use it to:

• Review recent account activity
• Summarize the deal history
• Identify open tasks
• Retrieve recent notes
• Review prior conversations
• Ask follow-up questions about the record

This reduces the time required to reconstruct the account history. The salesperson still determines the meeting objective and discovery strategy.

Sales Workspace Insights Help Representatives Prepare

Eligible users can review meeting attendees, recent activity, prior pain points, webpage visits, and Prospecting Agent interactions in the sales workspace.

A recent visit to a pricing page or a pain point raised during an earlier conversation may meaningfully change how a representative approaches the meeting.

Some meeting-preparation and follow-up capabilities are part of the Meeting Notetaker with Smart Deal Progression beta. Managing meetings in the sales workspace requires an assigned Sales seat, while Notetaker access depends on the applicable seat and beta configuration. HubSpot documents the current requirements here.

Smart Deal Progression Recommends Next Steps

Meeting Notetaker with Smart Deal Progression can analyze meeting transcripts and related deal information to recommend:

• CRM property updates
• Follow-up email drafts
• Tasks
• Next steps

The salesperson reviews and applies the recommendations. Custom properties added to smart data capture can consume HubSpot Credits, while HubSpot’s standard properties do not.

This addresses a common source of CRM deterioration: the period immediately after a meeting, when the representative moves to the next call before updating the record or sending follow-up.

Because AI can misinterpret a tentative comment as a firm commitment, the representative should confirm what the buyer agreed to before accepting the recommended updates. HubSpot currently identifies Smart Deal Progression as a beta.

5. How Does HubSpot AI Maintain and Analyze CRM Information?

Data Agent and Smart Deal Progression can turn research and conversations into structured CRM fields. Conversation Intelligence helps teams examine what happened across sales calls.

Conversation Intelligence makes recorded conversations searchable and can help surface recurring topics such as:

• Objections
• Competitor mentions
• Pricing concerns
• Common questions
• Customer pain points
• Coaching opportunities

This allows managers to review a broader sample of conversations than they could listen to manually. The analysis can point them toward calls worth examining, while the manager provides the context and coaching.

Organizations must also establish an appropriate legal basis for recording and processing conversations. Recording, consent, and disclosure requirements should be resolved before these capabilities are deployed.

Across all five jobs, the pattern holds: AI absorbs the repetitive research, drafting, and documentation work, while the salesperson keeps ownership of judgment, relationships, and accountability. That division of labor only holds up if the tools are actually set up to do their part.

What Does a Reliable Prospecting Agent Implementation Require?

Hypha organizes Prospecting Agent implementations into four connected layers:

Teams can activate these layers gradually, but the agent cannot be configured in isolation and expected to compensate for missing data, unclear targeting, or absent ownership.

A first Prospecting Agent pilot does not require every CRM property to be perfected. It does require a controlled group of eligible contacts, accurate company associations, clear audience-specific instructions, exclusion rules, and someone responsible for reviewing the results.

For a deeper explanation of why CRM context influences AI performance, read Is HubSpot Quietly Winning the AI Race?

Which HubSpot AI Tool Should Your Sales Team Start With?

Start with the sales job carrying the clearest administrative burden.

• Identify active accounts: Configure Buyer Intent around a narrow target market and a few meaningful behaviors, research topics, or company signals.
• Research and prioritize prospects: Test enrichment or Data Agent on a defined group of records and a question the team currently answers manually.
• Conduct outreach: Introduce Prospecting Agent in review mode with one audience, one value proposition, and a controlled contact group.
• Prepare for and follow up on conversations: Use Breeze Assistant and sales workspace insights for meeting preparation, or evaluate Smart Deal Progression if follow-up and CRM updates are routinely delayed.
• Maintain and analyze CRM information: Test Data Agent or Conversation Intelligence where incomplete fields or limited call visibility create a measurable problem.

Teams often gravitate toward Prospecting Agent because outreach is the most visible AI sales use case. In practice, incomplete account research or inconsistent post-meeting follow-up may provide an easier first implementation with a clearer baseline.

The initial deployment should be narrow enough to evaluate. Give one team one defined use case, document the existing process, and establish success measures before increasing volume or autonomy. Useful measures may include:

• Research time per account
• Draft-approval rate
• Percentage of emails requiring substantial edits
• Reply and meeting-booking rates
• Time from meeting completion to follow-up
• Completion of required CRM fields
• HubSpot Credits consumed per qualified opportunity

Track these before expanding a tool’s scope, adding a second use case, or increasing its autonomy—each tool needs responsibility for a specific job, not a general mandate to use more AI.

Putting HubSpot AI to Work

HubSpot AI can help sales teams identify active accounts, research and prioritize prospects, conduct outreach, support sales conversations, and maintain useful CRM data.

Its practical advantage comes from performing repetitive work with the customer context and sales history already stored in HubSpot. The salesperson remains responsible for deciding which opportunities matter, interpreting the buyer’s needs, and managing the relationship.

The best place to start depends on where the team is losing time. A company struggling with account research needs a different tool than one struggling with outbound capacity or post-meeting follow-up. Choosing the capability that matches the problem is more important than activating every AI feature available.

If Prospecting Agent is the piece your team is most likely to start with, Hypha’s outbound prospecting system covers how we configure the data, guardrails, and human review behind an implementation like this.

Frequently Asked Questions About HubSpot AI for Sales