Understand Breeze Assistant

Review how Breeze Assistant, HubSpot's conversational AI assistant, uses your account context to generate responses to your prompts. This article covers the context types Breeze Assistant uses, the available interaction methods, and what to expect from responses.

For an overview of configuring your account for HubSpot's AI, review how to set up a HubSpot account to adopt AI features.

Please note: you can manage AI feature access in your AI settings and configure what data is shared. Review HubSpot's AI Trust FAQs and AI Model Cards for detailed information on AI security controls, data use, and compliance.

Understand requirements

• Turn on Give users access to Breeze Assistant in your AI settings.

Understand limitations and considerations

• Breeze Assistant responses and actions depend on your user permissions. As a Super Admin, review user access to HubSpot AI tools.

Understand the context Breeze Assistant uses for responses

Breeze Assistant draws on three types of context when generating a response and may not use all three types each time. Depending on your account configuration, your permissions, and your specific request, Breeze Assistant uses varying levels of context. If a response seems incomplete or lacks key details, you can review the types below to determine what additional context to provide:

• Account-level business context: information configured in the context section of agent hub, such as your company profile, ideal customer profiles (ICPs), personas, and team processes like sales methodology. You can also create knowledge vaults that store more specific information for targeted use cases, such as product details or support guidance.
• Task-level context: information available for the specific request, such as the CRM record you're asking about, files you've added to a prompt, results from a connected app search, or knowledge vault content. Breeze Assistant accesses only information that's relevant to the request and available within your user's permissions.
• Personal and project context: information Breeze Assistant stores as memories, such as output preferences, tone, and recurring instructions. When you chat with Breeze Assistant within a project, those instructions, tools, and reference materials persist across conversations in the project.

Review ways to interact with Breeze Assistant

• Prompt: regular text-based input. Use regular prompts for research, general questions, and feature exploration.
• Saved prompt: regular prompts saved for repeat use. Save specific sets of instructions or template prompts to save time writing out instructions each time you chat with Breeze Assistant. You can also automate a saved prompt to automatically start a chat with Breeze Assistant on a recurring basis. For example, save a detailed prompt with instructions that resulted in a personalized email draft that required minimal editing before you sent it.
• Project: a folder of instructions, specific context, knowledge vaults, and tools. Start chats within a project and Breeze Assistant uses the saved set of materials. Use projects for conversations that share the same instructions, reference materials, or goals. For example, a project for a specific marketing campaign could include background documents, a target persona, and tone guidance that applies across every conversation in that project.
• Voice: voice-based conversation. Speak to Breeze Assistant, and it responds verbally. Use voice input when you want to share your thoughts aloud or be hands-free. For example, start voice mode after an in-person meeting with a client to save meeting notes to their contact record.

Understand the structure of a Breeze Assistant response

Each Breeze Assistant response varies based on your prompt and the context used. If a response appears to lack direction or not provide information you expect, review your context settings or iterate on your prompt. Based on your prompt, a response may include one or all of the following:

• Plain text output: answers to your questions or synthesis of research.
• Confirmation: a card displayed in the chat when a prompt requests CRM changes, with the option to confirm or cancel before the action runs.
• Follow-up questions: when more information is needed to complete your request, you may receive single, multiple choice, or free-text questions.
• Artifact: for eligible content types (e.g., email, report, spreadsheet) output appears as a saved and versioned piece of content.
• Sources: for prompts involving research, analysis, or data interpretation, citations appear inline or as footnotes in the response.

Use-case: use Breeze Assistant to analyze deals that need attention

A sales rep at Sprocket Supply Co. prompts Breeze Assistant: Review my open deals this week, identify the one at highest risk, and draft a re-engagement email for that contact. Breeze Assistant returns a plain text summary and generates a draft email artifact. Breeze Assistant uses the rep's saved email signature and connected calendar to add personalization to the drafted email. The rep notices that the deal record wasn't cited in the summary and prompts Breeze Assistant: always cite sources, regardless of if I asked about a specific CRM record. Then, the rep follows up on that prompt with Mark that deal as at-risk, and a confirmation card appears in the chat for manual approval before the Deal risk property is updated.

Additional resources

Use Breeze Assistant

• Use Breeze Assistant 
• Save prompts
• Summarize and review data
• Research HubSpot product updates
• Analyze your Breeze Assistant usage

Context and memory

• Manage AI context
• Build and manage knowledge vaults
• Import memories from third-party AI assistants

Build

• Create and manage projects
• Build custom AI agents with Breeze Assistant

Generate

• Generate data visualizations
• Generate documents and presentations

Mobile

• Use Breeze Assistant on mobile
• Use Breeze Assistant voice mode

Integrations

• Connect apps to Breeze Assistant