AD SPACE AVAILABLE · BRAND PARTNERS WELCOME

Configuring HubSpot Breeze Customer Agent to Resolve Support Tickets with Knowledge Base Documentation

365Finds Editorial
Indexed 11,430 Wiki articles

Put 365 Wiki to work with your AI

Copy the prompt below, paste it into the AI you already use, then ask your question.

Download full Skill

Try it with the prompt first. If you use 365 Wiki often, install the full 365Wiki Skill so you do not have to paste it again. No regular AI? Gemini or Microsoft Copilot on the web will work.

365Wiki Skill

When customer inquiry volume surges, support operations face a classic operational crunch: frontline agents spend hours answering repetitive inquiries—such as password resets, order tracking, returns, and subscription changes—while critical technical issues linger in the backlog. HubSpot Breeze Customer Agent (formerly Service Agent within Service Hub) addresses this bottleneck by autonomously parsing, answering, and resolving tickets using an organization's verified documentation, historical ticket records, and public resources.

Unlike rigid decision-tree bots that break when a user deviates from predetermined branching logic, Breeze Customer Agent leverages generative retrieval-augmented intelligence grounded in your company's actual data. This guide walks support managers, RevOps leads, and Service Hub administrators through configuring Breeze Customer Agent, mounting knowledge sources, mining historical tickets, and establishing foolproof escalation handoffs.


1. Prerequisites, Permissions, and the Billing Model

Before accessing the Customer Agent configuration panel, ensure your HubSpot portal meets the technical and administrative prerequisites:


2. Initial Setup: Agent Identity and Brand Persona

A support agent’s tone sets customer expectations from the opening greeting. Configuring identity correctly ensures brand alignment and clarity.

  1. In your HubSpot account, navigate to Service > Customer Agent (accessible directly or under the More menu).
  2. Click Continue to agent setup.
  3. Name Your Agent: Assign a recognizable name (e.g., "Sprocket Support Assistant" or "Acme Virtual Specialist"). Customers will see this name accompanied by a Powered by AI badge in chat interfaces.
  4. Set the Personality: Select a conversational demeanor from five standard profiles:
    • Friendly: Approachable and encouraging.
    • Professional: Concise, structured, and formal.
    • Casual: Relaxed, concise, and modern.
    • Empathetic: Validating, gentle, and understanding—ideal for dispute-heavy or sensitive services.
    • Witty: Lighthearted, suitable for direct-to-consumer lifestyle brands.
    • Use My Brand Voice: If your marketing or service teams configured an AI Brand Voice profile in HubSpot, select this option to mirror your global corporate messaging guidelines.
  5. Multi-Brand Portals: If your enterprise utilizes HubSpot Brands, you can create dedicated customer agents for each brand unit, configuring separate knowledge bases, URLs, and handoff queues for each division.

3. Mounting Knowledge Sources: Articles, Public URLs, and Files

The accuracy of Breeze Customer Agent depends strictly on the quality and structure of its knowledge sources. It only generates answers backed by verifiable source materials.

Navigate to Train > Knowledge in the left sidebar and click Add content to attach knowledge inputs:

A. HubSpot Native Knowledge Base & Web Pages

B. Public External URLs and Subdomains

If documentation lives on Notion, Zendesk, GitBook, or a custom CMS:

C. File Uploads and Internal Documentation

D. Direct Q&A "Short Answers"

For deterministic questions that demand an exact, approved script (such as holiday shipping cutoffs, return windows, or exact SLA tier definitions), click Create a short answer. Enter the precise question and answer. You can embed hyperlinks and illustrative diagrams directly within the editor.


4. Mining Historical Support Tickets to Generate Knowledge

Rather than building hundreds of FAQ entries manually, support operations can leverage HubSpot's built-in Generate knowledge feature. This tool uses AI to scan closed tickets and extract proven, agent-tested solutions.

  1. Under Train > Knowledge, click Add content > Generate knowledge.
  2. Choose the data origin:
    • Tickets: Select All tickets (which ingests closed tickets from the preceding 90 days) or select A specific ticket segment (e.g., tickets tagged with "Billing Inquiry" or "Tier 1 Resolved").
    • Inbox: Select All inboxes or target a specific support inbox.
    • File Upload: Upload batched transcript archives.
  3. Topic Guardrails: Toggle the Guardrails switch on.
    • In the Include topics box, enter high-frequency ticket drivers (e.g., "Account Lockouts", "Invoice Generation", "License Allocation").
    • In the Exclude topics box, enter subjects unsuitable for autonomous answers (e.g., "Executive Escalations", "Enterprise SLA Breaches", "Custom Contract Terms").
  4. Click Generate knowledge.
  5. Review and Approve Drafts:
    • Processing takes several minutes. Once complete, click Review drafts in the notification banner.
    • HubSpot clusters findings into structured topic cards displaying proposed Question & Answer pairs.
    • Click into any Q&A card to edit the generated response, refine terminology, or insert relevant links.
    • Click Accept to immediately convert the pair into an active knowledge source, Skip to defer, or Decline to discard.

5. Setting Communication Guidelines and Behavioral Boundaries

Guardrails prevent hallucinations and keep the AI strictly within support scope:

  1. Navigate to Train > Guidelines.
  2. Communication Style: Click + Add Communication style. Name the rule and attach it either globally or to specific contact segments (such as VIP clients versus Free tier users). Define formatting preferences (e.g., "Always structure troubleshooting steps in numbered lists; maintain concise answers under 120 words; do not speculate on unreleased roadmap features").
  3. Guardrails (Restricted Topics): Under the Guardrails block, explicitly declare topics the agent must never discuss. For example, if a user asks about legal disputes, competitor comparisons, or unannounced discounts, instruct the agent: "Inform the customer politely that this topic requires specialized review, and initiate a human handoff immediately."
  4. Custom Instructions: Provide contextual guidance for edge cases, such as handling emotionally frustrated customers with de-escalation validation before executing an agent transfer.
  5. Click Test in the upper-right corner to simulate customer inputs, then click Publish.

6. Configuring Ticket Escalation and Human Handoff Rules

An autonomous agent is only as safe as its handoff mechanics. When a customer reaches the agent's boundaries, the transfer to a human support engineer must be immediate, context-rich, and orderly.

Handoff Triggers

Navigate to Train > Human handoff. The Customer Agent triggers a transfer under three default conditions:

  1. Low Confidence: The agent cannot locate an answer in synced knowledge sources.
  2. Explicit Request: The customer types "agent", "human", "representative", or "support team".
  3. Paused State: The agent has been manually paused or credit allotments are exhausted.

To add custom business triggers, locate the Handoff guidelines field. Use declarative syntax or prebuilt templates:

"When the customer requests to speak with a human or uses any of the following triggers, you must immediately initiate a handoff: 'Cancel subscription', 'Issue refund', 'Security vulnerability', 'System down', 'Talk to billing'."

Configuring the Transfer Pipeline

Click Edit handoff process and choose the handoff execution style:

Routing Destinations: Direct vs. Workflow Rules

Understanding Ticket Status Progression

During handoff, HubSpot automatically updates the CRM property Customer Agent ticket status:

Status ValueSystem StateOperational Action Required
PendingAgent is actively conversing with the customer; no handoff triggered.Reps do not need to intervene.
Awaiting Human AssignmentHandoff triggered; ticket moved to Help Desk, but awaiting an available rep or workflow assignment.Monitor queue; reps can manually claim unassigned tickets.
Handed offTicket successfully transferred and assigned to an active human representative.Human rep takes ownership; reviews AI conversation history.

Availability and Messaging


7. Pre-Deployment Testing and Phased Rollout

Never deploy an AI agent to 100% of live traffic on day one. Validate accuracy and operational guardrails through structured testing.

1. Interactive Preview and Testing Insights

2. Batch Testing

Run automated batch evaluations by feeding historical inquiries to the agent at scale. Review automated deflection scores, citation accuracy, and flag hallucinations before launching.

3. Deploying to Support Channels with Traffic Throttling

  1. Navigate to Deploy > Channels and click Deploy agent.
  2. Select your target channel (e.g., Website Live Chat or Support Email).
  3. Set Working Hours: Choose All hours, During specific hours, or Only outside specific hours (ideal if using the agent solely for overnight deflection).
  4. Conversation Coverage Percentage: Enter a conservative coverage threshold (e.g., start at 15% to 25%). HubSpot randomly routes that proportion of incoming sessions to Breeze, letting the remainder flow to standard human queues.
  5. For email channels, filter by Contact lists to include / exclude to test on non-critical customer segments first.
  6. Click Deploy.

8. Post-Launch Monitoring and Knowledge Gap Iteration

After deployment, review operational dashboards under Service > Customer Agent > Analyze:

By systematically curating knowledge sources, mining historical tickets, enforcing strict guardrails, and linking escalation rules to Service Hub workflows, support teams transform high-volume ticket surges from a frontline crisis into a scalable, high-satisfaction resolution engine.


References