How to set up HubSpot Service Hub AI: A complete guide for 2026

Setting up AI-powered customer support can feel overwhelming. You're essentially training a digital team member to represent your company 24/7. Get it right, and you'll free up your human agents for complex issues while customers get instant answers. Get it wrong, and you'll risk frustrating customers with unhelpful responses.

HubSpot Service Hub includes a powerful AI tool called Breeze Customer Agent that can handle customer conversations across multiple channels. This HubSpot Service Hub AI setup guide walks you through the complete setup process, from initial configuration to going live.

What you'll need before starting

Before diving into the setup, make sure you've got these prerequisites in place:

• Service Hub Professional or Enterprise subscription. The Breeze Customer Agent is not available on Free or Starter plans. Professional starts at $90 per seat per month (annual billing), while Enterprise is $150 per seat per month.
• Super Admin permissions. You'll need access to your HubSpot account's AI settings to enable the features.
• At least one connected channel. This could be live chat, Facebook Messenger, WhatsApp, or email connected to your Conversations Inbox or help desk workspace.
• Content sources ready. Have your knowledge base articles, help docs, or website pages prepared. The AI will learn from these to answer customer questions.
• HubSpot tracking code (for external sites). If you're deploying the agent on a website not hosted on HubSpot, you'll need the tracking code installed.

If you're exploring AI support options beyond HubSpot's native tools, we also offer HubSpot integration that connects additional knowledge sources to your existing workflow.

Step 1: Enable AI features in your HubSpot account

First, you need to turn on the AI capabilities at the account level. This gives your team access to the generative AI tools and configures what data sources the AI can reference.

Navigate to your HubSpot settings by clicking the gear icon in the top navigation bar. In the left sidebar, find Account Management and select AI. You'll see several toggles that control AI access:

Toggle on "Give users access to generative AI tools and features." This is the master switch that enables AI functionality across your account. Without this, nothing else'll work.

Next, configure your data access permissions. These determine what information the AI can use when generating responses:

• CRM data: Allows the AI to access customer records, deal history, and contact properties. This is essential for personalized responses.
• Customer conversation data: Lets the AI learn from past tickets, emails, and chat transcripts to improve responses.
• Files data: Enables the AI to reference files stored in your HubSpot file manager.

For teams looking to expand their AI capabilities beyond HubSpot's native features, our AI chatbot offers additional flexibility for customer-facing conversations.

While you're in the AI settings, take a moment to configure your brand voice. Navigate to the Data sources tab and set up your Brand Kit, Company Profile, and Ideal Customer Profile. These help the AI understand your company's tone, industry, and target audience, which improves response quality significantly.

Step 2: Create your Breeze Customer Agent

Now that AI features are enabled, it's time to build your actual customer agent. Think of this as creating the personality and knowledge base for your digital support rep.

Go to Service → Customer Agent in your HubSpot account. Click "Set up your agent" in the upper right corner.

Start by giving your agent a name. Choose something descriptive that your team'll recognize, like "SupportBot" or "Help Assistant." This name appears internally and helps you identify the agent in reports.

Next, select a personality from the dropdown. Your options are:

• Friendly: Warm and approachable, good for consumer brands
• Professional: Polished and business-like, ideal for B2B companies
• Casual: Relaxed and conversational, works well for startups and modern brands
• Empathetic: Understanding and patient, great for sensitive industries
• Witty: Light and humorous, suitable for brands with playful voices

If you've already configured brand voice in your AI settings, you can also select "Use my brand voice" to have the agent adopt your custom tone.

Click Next to move to the Content page. This is where you teach your agent what it needs to know.

Step 3: Configure knowledge sources and training

Your agent is only as good as the info you give it. HubSpot lets you connect multiple types of content sources to train your AI.

Start with the "Add existing HubSpot content" dropdown. Select the types of content you want the agent to learn from:

• Knowledge base articles: Your help documentation and FAQs
• Website pages: Product pages, about pages, and other informational content
• Landing pages: Campaign-specific pages with detailed information
• Blogs: Educational content that answers common questions

For external content, enter public URLs in the designated field. If you want the agent to learn from an entire domain, check the "Import related URLs" box. This automatically crawls and syncs all pages within that domain.

You can also upload files directly. Supported formats include .docx, .pdf, .csv, .xls, .html, .txt, .json, .md, and many others. CSV and XLS files work particularly well for structured data like product catalogs and pricing information.

A few best practices for selecting content:

• Start with your most frequently asked questions and the content that answers them
• Use clear headers and subheaders in your knowledge base articles (the AI uses these to understand content structure)
• Include bullet points and lists for easier readability
• Avoid uploading sensitive information or private data
• Keep content updated regularly, the AI learns from what you provide

When you've added all your content sources, click "Create agent" in the bottom right. The agent'll begin processing your content, which typically takes a few minutes depending on the volume.

Step 4: Assign your agent to channels

With your agent created and trained, you now need to tell it where to work. HubSpot lets you deploy the agent across multiple channels.

Navigate to Service → Customer Agent, then click the Manage tab. In the left sidebar, select Deployment → Channels. Click "Deploy agent" in the top right.

Select the channel you want to assign the agent to from the dropdown. Your options include:

• Live chat on your website
• Facebook Messenger
• WhatsApp
• Email
• Calling (currently in beta)
• Custom channels

Next, configure working hours. You have three options:

• All hours: The agent responds to conversations 24/7
• During specific hours: Set time zones, days, and hours when the agent should be active
• Only outside specific hours: The agent covers nights, weekends, or other off-hours

If you choose specific hours, you'll also need to configure what happens when the agent isn't available. Typically, you'd route these conversations to human agents or leave 'em unassigned for your team to handle during business hours.

Click "Deploy" to activate the agent on that channel. You can repeat this process to deploy the same agent across multiple channels.

Step 5: Set up handoff and escalation rules

Not every conversation should stay with the AI. You need clear rules for when the agent should transfer to a human.

HubSpot calls this "handoff." You can create custom triggers that automatically transfer conversations based on specific conditions.

Common handoff scenarios include:

• The AI can't answer the question after multiple attempts
• The customer specifically requests a human agent
• The conversation involves sensitive topics like billing disputes or cancellations
• The customer uses keywords you've flagged for human review

To configure handoff, go to your Customer Agent settings and look for the handoff configuration section. You can set up rules that route conversations to specific users, teams, or leave them unassigned for anyone to pick up.

You can also use ticket-based workflows to create more complex routing logic. For example, you might route VIP customers to a dedicated support team while sending general inquiries to a shared queue.

Important note: Only paid Sales Hub or Service Hub users can be used in routing rules. Make sure your team members have the appropriate seats assigned.

Step 6: Test and publish your AI agent

Before going live, thoroughly test your agent. HubSpot provides a preview mode that lets you simulate conversations without consuming credits or affecting real customers.

Navigate to Service → Customer Agent and click "Test [agent name]" in the top right corner.

Select the channel you want to test (Email or Live Chat). You can either use pre-populated test questions or enter your own custom questions.

As you test, pay attention to:

• Response accuracy: Does the agent answer correctly based on your knowledge base?
• Tone consistency: Does it sound like your brand?
• Source citations: Is it properly citing where information comes from?
• Handoff triggers: Do escalation rules work as expected?

The Testing Insights panel on the right shows you why the agent responded a certain way, which sources it referenced, and what triggers were activated. This is invaluable for debugging any issues.

If you get a response that isn't quite right, click "Improve response" and use the dropdown to adjust how the agent should handle similar questions in the future. You can add handoff triggers, create shorter answers, or flag knowledge gaps.

Once you're satisfied with the testing, toggle the agent to active. It will now respond to real customer conversations on your assigned channels.

Tips for optimizing your HubSpot Service Hub AI setup

Getting the agent live is just the beginning. Here are ways to improve performance over time:

• Monitor unanswered questions. The Performance tab shows you what customers are asking that the agent couldn't answer. Use this to identify content gaps in your knowledge base.
• Structure content for AI. Use clear headings, short paragraphs, and bullet points in your knowledge base articles. The AI parses this structure to find relevant information.
• Update content regularly. Outdated information leads to incorrect answers. Set a schedule to review and refresh your knowledge sources.
• Use the Breeze Knowledge Base Agent. This companion tool automatically drafts new articles based on support ticket data, helping you fill content gaps faster.
• Review conversation transcripts. Regularly check how the agent is handling real conversations and adjust your training content accordingly.

For a deeper dive into HubSpot AI capabilities and integration options, check out our complete HubSpot Service Hub AI integration guide.

Extending HubSpot AI with eesel AI

HubSpot's native AI is powerful, but it's got limitations. The Breeze Customer Agent primarily learns from HubSpot-hosted content and public URLs. If your knowledge is spread across multiple platforms, you might find the native solution limiting.

This is where we can help. Our AI agent integrates with HubSpot while connecting to knowledge sources across your entire tech stack:

• Unified knowledge: We pull from Confluence, Google Docs, Notion, PDFs, and other sources, not just HubSpot content
• Advanced simulation: Test your AI on thousands of past tickets before going live, so you know exactly how it will perform
• Predictable pricing: Our pricing is based on interactions, not credits, making costs easier to forecast
• Cross-platform consistency: Deliver the same accurate answers whether customers reach you through HubSpot, Slack, or other channels

Think of HubSpot's native AI as a solid foundation, and our integration as the expansion pack that connects your whole knowledge ecosystem.

Get started with AI-powered customer support

Setting up HubSpot Service Hub AI takes some upfront work, but the payoff's significant. Teams using the Breeze Customer Agent resolve 65% or more of conversations automatically, with top performers hitting 90%. That translates to faster response times, happier customers, and less burnout for your human agents.

Start with the steps outlined above, test thoroughly before going live, and continuously refine based on performance data. And if you need to connect knowledge sources beyond HubSpot, we're here to help extend your AI capabilities.

Ready to explore what's possible? See how we integrate with HubSpot to create a more comprehensive AI support solution.