Introduction: Transforming Communication with AI

HubSpot's AI Customer Agent, powered by Breeze AI, represents a significant leap forward in automated customer support— including turning your website into a responsive, helpful, and continuously improving front-of-house assistant.

This comprehensive guide will walk you through (just about) every aspect of setting up, training, and optimizing your HubSpot AI Customer Agent. In turn, it’ll help you to deliver exceptional customer experiences… while freeing your human agents to focus on complex, high-value interactions.

What is HubSpot’s AI ‘Customer Agent’?

HubSpot AI Customer Agent is an intelligent conversational assistant that uses your existing HubSpot content and public URLs to provide contextual, accurate responses to customer inquiries.

This AI-powered solution:

• Responds to support questions with a conversational tone
• Provides answers backed by verifiable sources
• Automatically detects and responds in the customer's browser language
• Seamlessly hands off complex queries to human agents when needed
• Continuously learns from your content to improve response accuracy

And here’s a great overview from HubSpot's very own…Kyle Jepsen:

Key Benefits

Here’s just a handful of benefits you'll experience once up and running…

Enhanced Customer Experience

• 24/7 availability for instant response
• Consistent, accurate information delivery
• Multilingual support capabilities
• Contextual responses based on your specific content

Operational Efficiency

• Reduces workload on human support agents
• Handles routine inquiries automatically
• Provides detailed source attribution for transparency
• Integrates seamlessly with existing HubSpot workflows

Scalable Support

• Handles multiple conversations simultaneously
• Adapts to your growing content library
• Maintains consistency across all interactions
• Improves over time through AI learning (and human inputs!)
  

Now you're warmed up and ready to go, let's look at the step-by-step guide…

Step-by-Step Setup Guide

Pre-planning:

Before you even get started, you'll need to toggle on a couple of switches in the settings area:
In your HubSpot account, click the settings settings icon in the top navigation bar:
In the left sidebar menu, in the Account management section, navigate to AI.
Then you'll need to toggle these two settings (if not already done):

Full resource and further information here.

NOW you can get started…

Phase 1: Creating Your Customer Agent

Navigate to the Customer Agent Dashboard

• In your HubSpot account, go to Service > Customer Agent
• Click "Launch" to access the Customer Agent interface
• Select "Create Agent" to begin the setup process

Configure General Information

Agent Name Choose a descriptive name that reflects your agent's purpose (e.g., "Support Assistant," "Product Helper," or your company name + "Bot")

Role Selection Select the most appropriate role for your agent:

• Customer Support: Best for general inquiries, troubleshooting, and FAQ responses
• Marketing Specialist: Ideal for product information, feature explanations, and lead qualification
• Sales Representative: Optimized for pricing questions, demo requests, and initial sales conversations

Personality Configuration Choose a personality that aligns with your brand voice:

• Friendly: Warm, approachable, and conversational
• Professional: Formal, business-focused, and authoritative
• Casual: Relaxed, informal, and approachable
• Empathetic: Understanding, supportive, and caring
• Witty: Clever, engaging, and memorable

Pro Tip: you can configure brand voice settings in HubSpot (under the "Branding" tab within "Account Defaults") for the whole business:

Then you can configure:

More on brand voice here.

Phase 2: Content Configuration

Adding HubSpot Content Sources

As your Customer Agent is powered by content, you really want to make sure that what you feed into the system is what you want it to serve up e.g.

Knowledge Base Articles

• Select your most comprehensive and up-to-date FAQ articles
• Ensure articles use clear headers and subheaders
• Include articles that address common customer pain points

Website and Landing Pages

• Add key product pages that explain features and benefits
• Include service description pages
• Add pricing and plan information pages

Blog Content

• Select evergreen blog posts that answer common questions
• Include how-to guides and tutorials
• Add industry insights that position your company as an expert

Product and services information

• Up-to-date pricing information
• Technical spec
• Approach to delivery of services etc

As you can see,  there are many options, and personally, I tend to favor files of 'FAQs' as a starting point.
(Unless you have a well-manicured knowledge base, for instance)

You can add these here…

Supported File Types You can upload various file formats containing relevant information:

• PDF documents (product guides, manuals, policies)
• Word documents (FAQs, internal guides)
• Text files (scripts, common responses)

And a few tips…

• Use descriptive file names
• Ensure content is current and accurate
• Remove any sensitive or confidential information
• Structure content with clear headings and bullet points

Very important…

You need to decide whether you want the citations of the answers to be ‘viewable’.
If you toggle
In other words, people won't just see the answers, they'll see the source of the answer as well.

If you toggle, then you make the files ‘available to be cited’ in answers - which is exactly what we've done with our ISO 27001 and ISO 9001 certificates:

Note: you may not want this to be the case across all of your content.

Once you have your content ready, you can move to the next phase…

Phase 3: Channel Assignment

Select Your Deployment Channel:

Inbox vs. Help Desk Configuration Choose whether to connect through:

• Conversations Inbox: Best for general customer communications
• Help Desk Workspace: Ideal for structured support ticket management

Channel Selection Assign your agent to appropriate channels:

• Live Chat: Website visitors get immediate assistance
• Facebook Messenger: Social media customer support
• WhatsApp Business: Mobile-first customer communication
• Email (BETA): Automated email response handling

Configure Handoff Rules:

Handoff to Human Agent Configure when and how conversations transfer to human agents:

• Select specific users or teams for escalation
• Set up routing rules based on inquiry complexity

Stay Assigned Option Choose this if you want the AI to remain engaged even when it can't answer:

• Customize the "handoff unavailable" message
• Maintain conversation continuity
• Allow for follow-up questions and clarification

Testing and Optimization

Next, it's time for some ‘Pre-Launch Testing’
Before going live, test your agent with:

• Common customer questions from your support history
• Edge cases and unusual inquiries
• Questions that might reveal knowledge gaps
• Multi-part or complex questions

See how it responds, and you may need to adjust the source of the information, and/or use short answers to fill in any gaps:

Advanced Configuration Strategies

You can also set up something called Actions, which is definitely a more advanced approach where you can…

“Empower Customer Agent to perform tasks just like your support reps. Configure actions like sending invoices, sharing order status, or sending password reset links — all without the hassle of manual intervention.”

More information on that here. 

Once you GO LIVE…

When your agent frequently expresses uncertainty, the issue is usually poor content quality or gaps in your knowledge base. 

If you run the test yourself and then review the responses, you'll often find the source of the issues. 

This is why ‘Short Answers’ (mentioned above) are SO useful…
Spot a ‘Knowledge gap’ once live (By going to this tab below)

Click on the hyperlink relating to the gaps, and then use short answers to build out the information to be used by the Customer Agent, by clicking on the ‘Create an answer’ button:

And then add in both the question and the answer. 

You can even test it before you go live:

This gives a way for your customer agent to build a micro knowledge base through ‘short answers’. 
It's early days with our own experimentation, but I'm extremely excited with the potential. 

Conclusion: Your Journey to AI-Powered Customer Success

Setting up and optimizing your HubSpot AI Customer Agent is just the beginning of your journey toward enhanced customer support and operational efficiency. Remember that this is an iterative process—continuous testing, monitoring, and optimization will ensure your agent becomes increasingly effective over time.

The key to success lies in:

• Starting with high-quality, comprehensive content
• Regular monitoring and optimization
• Maintaining alignment with your brand voice and values
• Focusing on customer experience outcomes

As you implement these strategies, you'll find that your AI Customer Agent becomes an invaluable team member, handling routine inquiries with precision while enabling your human agents to focus on complex, high-value customer interactions.

Keep checking and testing that the right content appears in conversations, and remember—this is where your journey to exceptional AI-powered customer support truly begins.