HubSpot Breeze AI Agent Workflows: 2026 Guide
HubSpot shipped 11 platform updates on January 27, 2026, headlined by the GPT-5 migration for Breeze Studio agents, customer message insights with audit cards, and the new Run Agent workflow action. This guide covers what changed, what it means for your CRM automation strategy, and how to configure agent-powered workflows across marketing, sales, and service hubs.
Updates Released Jan 27
Default LLM Upgrade
Supported Channels
Credits per Conversation
Key Takeaways
January 2026 Breeze AI Updates
HubSpot's January 27, 2026 release delivered a focused set of upgrades across customer agent capabilities, Help Desk functionality, workflow automation, and CRM visualization tools. The updates collectively address three priorities: trust through audit trails, flexibility through workflow integration, and extensibility through developer tools.
The headline changes center on Breeze AI agents. The GPT-5 model migration for Breeze Studio agents landed on January 12, followed by the customer message insights, Run Agent workflow action, and expanded omnichannel support in the January 27 batch. Together, these updates move Breeze from an isolated AI feature set toward a comprehensive agent infrastructure that integrates with the full HubSpot automation stack.
Full January 27 Update Summary
GPT-5 Migration for Breeze Studio Agents
According to HubSpot's developer changelog, Breeze Studio agents now default to GPT-5 instead of GPT-4.1. This migration affects marketplace agents and custom agents in private beta. Existing agents created before December 1, 2025 were automatically upgraded as of January 12, 2026.
The upgrade is selective. Customer, Prospecting, and Data Agents are excluded from the GPT-5 migration and continue operating on their current models. This distinction matters for teams planning automation around specific agent capabilities, as the model backbone directly affects response quality, reasoning depth, and action planning.
Which Agents Got GPT-5
| Agent | Model | Status | Type |
|---|---|---|---|
| Deal Loss Agent | GPT-5 | Auto-upgraded | Marketplace |
| Customer Health Agent | GPT-5 | Auto-upgraded | Marketplace |
| Customer Handoff Agent | GPT-5 | Auto-upgraded | Marketplace |
| Social Post Agent | GPT-5 | Auto-upgraded | Marketplace |
| Custom Agents (Beta) | GPT-5 | Auto-upgraded | Private Beta |
| Customer Agent | Unchanged | Not affected | Core |
| Prospecting Agent | Unchanged | Not affected | Core |
| Data Agent | Unchanged | Not affected | Core |
The GPT-5 backbone brings improved reasoning, better context retention across longer conversations, and more reliable action planning for multi-step agent tasks. For AI transformation initiatives, the model upgrade means agent-driven workflows can handle more complex decision trees without degrading output quality.
Customer Message Insights and Audit Cards
The audit card feature, now live for all accounts with the Breeze Customer Agent, delivers transparency into what the agent does during conversations. When the Customer Agent identifies a customer, qualifies a lead, or modifies CRM properties, an audit card appears showing the specific actions performed.
This is a direct response to the primary objection teams raise when deploying AI agents: uncertainty about what the agent actually does behind the scenes. Audit cards convert that uncertainty into verifiable records, which is critical for compliance-sensitive industries and teams scaling agent deployment beyond pilot programs.
What Audit Cards Display
Practical Impact on Support Operations
Before audit cards, support managers had to manually review conversation transcripts and cross-reference CRM records to verify agent behavior. Now, every agent action generates a timestamped card visible in the conversation timeline. This cuts audit time significantly and gives supervisors the ability to spot-check agent performance at scale.
- Quality assurance: Review agent actions without reading full transcripts. Filter audit cards by action type (property changes, qualifications, handoffs) to focus QA efforts on high-impact interactions.
- Training feedback: Identify patterns where the agent makes incorrect qualification decisions or sets wrong property values. Use these patterns to refine agent instructions in Breeze Studio.
- Compliance documentation: Audit cards serve as records of AI-driven CRM changes. For regulated industries, this paper trail demonstrates that automated actions were logged and reviewable.
- Escalation context: When the agent hands off to a human rep, the audit card summarizes everything the agent already did, preventing the customer from repeating information.
Set up a weekly audit review cadence during the first 30 days after enabling the Customer Agent. Focus on conversations where the agent qualified leads differently than your team would have. Adjust the agent's qualification criteria in Breeze Studio based on these discrepancies. Most teams find the agent stabilizes to acceptable accuracy within 2-3 rounds of instruction refinement.
Run Agent Workflow Action
The Run Agent workflow action, currently in private beta, addresses a limitation that constrained Breeze adoption since launch: agents were confined to Breeze Studio and could not be triggered by external events. The new action lets teams embed agent execution inside any HubSpot workflow, connecting AI agent reasoning to the full automation stack.
This is where Breeze shifts from a standalone AI tool to an automation primitive. Instead of agents waiting for users to invoke them, workflows can trigger agents based on any CRM event: deal stage changes, form submissions, lifecycle stage transitions, ticket creation, or scheduled cadences.
Configuration Options
- Choose from marketplace agents or custom agents built in Breeze Studio
- Configure agent-specific instructions for the workflow context
- Chain multiple agents in sequence for complex multi-step processes
- Pass enrolled record properties as agent input context
- Include associated records (company, deal, ticket) for richer context
- Map agent output fields to workflow branches and downstream actions
Example Workflow Patterns
Lead Qualification on Form Submission
Trigger: Contact submits a demo request form. Action: Run the Customer Health agent to research the company, score fit against ICP criteria, and update the lead status property. Branch: Route qualified leads to sales, partially qualified to nurture, and unqualified to a polite decline email.
Deal Risk Assessment on Stage Change
Trigger: Deal moves to negotiation stage. Action: Run the Deal Loss agent to analyze deal history, engagement patterns, and comparable closed-lost deals. Output: Risk score property updated on the deal record. Branch: High risk deals trigger manager notification and mandatory review meeting task.
Churn Prevention on Ticket Escalation
Trigger: Support ticket escalated to Tier 2. Action: Run the Customer Health agent to assess overall account sentiment from recent tickets, NPS scores, and usage data. Output: Health score updated on the company record. Branch: At-risk accounts trigger customer success manager outreach.
Breeze Agents: Features by Hub
HubSpot organizes Breeze agents around its three core hubs: Marketing, Sales, and Service. Each hub has dedicated agents optimized for its workflows, plus shared infrastructure through Breeze Studio and the Data Agent. Understanding which agents map to which hub is essential for planning your marketing automation strategy and CRM investment.
Core Breeze Agents
Drafts marketing materials by analyzing existing content and CRM data to match your brand voice. Generates blog posts, email copy, social posts, and landing page content.
- Brand voice matching from existing assets
- Multi-format output (email, blog, social)
- CRM data integration for personalization
Automates outbound sales by researching contacts, building company profiles from domain data, and generating personalized email sequences.
- Contact research and enrichment
- ICP and value proposition inference
- Personalized sequence generation
Conversational support agent trained on your knowledge base, website content, and help articles. Handles customer inquiries across nine supported channels.
- Trains on PDFs, KB articles, 1,000+ pages
- Personalized actions (order status, resets)
- Omnichannel: chat, email, WhatsApp, SMS
Identifies gaps in your documentation by reviewing support tickets and drafting help articles to fill missing coverage areas.
- Ticket analysis for documentation gaps
- Automated article draft generation
- Content quality recommendations
Data Agent Workflow Actions
The Data Agent operates across all hubs and provides three workflow-specific actions for processing CRM data at scale. These actions consume 10 credits per execution and can be combined with the Run Agent action in complex workflows.
| Action | Function | Use Case |
|---|---|---|
| Custom Prompt | Analyze, summarize, or categorize enrolled record data using a custom AI prompt | Sentiment analysis on support tickets, lead scoring commentary |
| Research | Research a company or contact using web data and domain analysis | Enriching new leads, competitive intelligence on deal contacts |
| Fill Smart Property | Populate CRM properties with AI-generated data based on record context | Auto-filling industry, company size, or technology stack fields |
Pricing, Credits, and Plan Requirements
HubSpot's Breeze pricing model has shifted toward transparent, usage-based billing. Features consume credits from a monthly allocation included with your subscription tier. Understanding the credit economy is essential for budgeting agent deployments and avoiding unexpected costs.
Credit Consumption by Feature
| Feature | Credits Used | Billing Unit |
|---|---|---|
| Customer Agent | 100 credits | Per conversation |
| Prospecting Agent | 100 credits | Per monitored contact/month |
| Data Agent | 10 credits | Per research prompt |
| Breeze Workflow Actions | 10 credits | Per action execution |
| Buyer Intent | 10 credits | Per company created/monitored |
| Standard Data Enrichment | Free | Included with Core Seat |
Plan Requirements for Breeze Agents
Billed annually. Content Agent + Data Agent workflow actions for email personalization.
Billed annually. Prospecting Agent + Deal Loss Agent + selling profile generation.
Billed annually. Customer Agent + Knowledge Base Agent + Help Desk blocking.
For additional credits beyond your included allocation, Breeze Intelligence credits for data enrichment are available at $45/month for 5,000 credits. The quantity of included monthly credits varies based on the highest-level product edition in your subscription and credits are not additive across multiple subscriptions.
Practical Workflow Automation Templates
The following templates combine Breeze agents with standard HubSpot workflow actions to create end-to-end automation sequences. Each template specifies the trigger, agent action, output handling, and estimated credit consumption per execution.
Marketing Hub: Lead Scoring and Nurture Routing
Trigger: Contact fills out content download form
Form submission enrollment triggers the workflow for all new leads from gated content.
Data Agent: Research company domain
Enrich the contact record with company data, industry classification, and estimated revenue. Cost: 10 credits.
Data Agent: Custom prompt for ICP scoring
Score the enriched contact against your ideal customer profile criteria. Update a custom lead score property. Cost: 10 credits.
Branch: Route by score
High score: Assign to sales rep. Medium: Enroll in 5-email nurture sequence. Low: Add to newsletter list only.
Total cost per lead: 20 credits | Monthly at 500 leads: 10,000 credits
Sales Hub: Deal Velocity Monitoring
Trigger: Deal sits in same stage for 14+ days
Scheduled workflow checks deal stage dates daily and enrolls stalled deals.
Run Agent: Deal Loss agent analyzes risk
Agent reviews engagement history, email open rates, meeting frequency, and comparable deals. Cost: 10 credits.
Output: Update deal risk score and recommended next action
Agent populates a custom risk score property and writes a recommended action note on the deal timeline.
Branch: Notify or escalate
High risk: Slack notification to sales manager + create task. Medium: Email reminder to deal owner with agent recommendations.
Total cost per deal check: 10 credits | Monthly at 200 stalled deals: 2,000 credits
Service Hub: Intelligent Ticket Routing
Trigger: New support ticket created
All incoming tickets from email, chat, or form submission enter the workflow.
Customer Agent: Attempt automated resolution
Agent searches knowledge base, provides answer, and asks if the issue is resolved. Cost: 100 credits per conversation.
Branch: Resolved or escalated
Resolved: Close ticket with satisfaction survey. Unresolved: Route to appropriate Tier 1 queue based on agent's category classification.
Audit card: Record agent actions
Audit card logs what the agent attempted, which KB articles were referenced, and why the issue was or was not resolved.
Total cost per ticket: 100 credits (if agent engages) | Monthly at 1,000 tickets: 100,000 credits
Strategic Recommendations for 2026
The January 2026 updates signal HubSpot's direction: agents as composable automation building blocks rather than standalone tools. Teams that treat Breeze agents as workflow primitives alongside standard actions like delays, branches, and property updates will extract the most value. Here is a structured approach to planning your deployment.
- Identify workflows with manual decision points that agents could automate
- Map current lead scoring rules that could be replaced by Data Agent analysis
- Calculate credit consumption for your monthly workflow volume to budget accurately
- Deploy agents first on high-value workflows with lower enrollment counts to control credit costs
- Deal risk assessment on enterprise deals, not every contact form submission
- Measure agent accuracy for 30 days before expanding to higher-volume workflows
- Establish a weekly review cadence for agent audit cards during pilot phase
- Track agent qualification accuracy against human decisions as a benchmark
- Document audit findings for compliance teams in regulated industries
- Prioritize channels by customer volume: deploy Customer Agent on your highest- traffic channel first
- Train the agent on channel-specific FAQs, not just generic knowledge base content
- Test WhatsApp and SMS channels separately, as message length limits affect agent response quality
Conclusion
The January 2026 Breeze updates mark HubSpot's transition from offering AI agents as standalone features to embedding them as composable primitives in the workflow engine. The GPT-5 migration improves reasoning quality for Breeze Studio agents. Audit cards solve the trust problem that held back agent adoption. The Run Agent workflow action removes the isolation constraint that limited agent utility.
For teams already on HubSpot Professional or Enterprise tiers, the practical next step is to identify one high-value, low-volume workflow where an agent can replace manual decision-making. Deploy it, measure accuracy using audit cards for 30 days, then expand based on results. The credit-based pricing model rewards targeted deployment over broad rollouts, so starting small is both the prudent and cost-effective approach.
Ready to Automate Your CRM with AI Agents?
Our team helps businesses configure HubSpot Breeze agents, design workflow automation strategies, and measure AI-driven CRM performance across marketing, sales, and service operations.
Frequently Asked Questions
Related CRM & Automation Guides
Continue exploring AI-powered CRM automation and marketing workflows