🔷 HubSpot Sales Practical Textbook — 2026 Edition
Chapter 11

Breeze AI and Sales Hub
Sales productivity with AIinnovate

The heart of HubSpot's AI strategy is Breeze is.With a three-layer structure of Breeze Copilot (AI assistant), Breeze Agents (autonomous AI agents), and Breeze Intelligence (data enrichment), it automates repetitive tasks for salespeople such as research, email creation, data entry, and prospect management, creating an environment where humans can focus on making decisions and building relationships.2025 is the year when the use of AI in Sales Hub undergoes a qualitative transformation, including the full-scale operation of Prospecting Agent, real-time detection of Deal Risk, and AI integration with CPQ.This chapter explains all the features of Breeze AI and how to use it practically.

📖 Estimated reading time: 35 minutes
🎯 Target audience: Sales representatives, sales managers, RevOps
📅 March 2026 version — Breeze Prospecting Agent officially operational

📋 Contents of this chapter

  1. 11-1Three-layer structure of Breeze AI: Overview of Copilot, Agents, and Intelligence
  2. 11-2Configuring and operating Breeze Prospecting Agent
  3. 11-3Designing the Selling Profile: Giving the “context” for AI to work correctly
  4. 11-4Limitations of leveraging and personalizing AI email generation
  5. 11-5Automated Conversation Intelligence and Deal Risk Detection
  6. 11-6AI credit management and ROI thinking
Section 11-1

Three-layer structure of Breeze AI: Overview of Copilot, Agents, and Intelligence

HubSpot's AI capabilities are integrated under the Breeze brand. More than just a chatbotOperates after understanding the sales context by combining CRM data, conversation history, and external informationIt is characterized by points. By accurately understanding the three-layer structure, you can quickly determine which function will solve your problem.

🤖 Breeze AI — 3-tier structure and key features in Sales Hub
💬
Layer 1 — Copilot
Breeze Copilot
HubSpot is an AI assistant that resides on the right side of your screen. Instantly perform tasks such as ``draft an email to this contact'', ``summary of this deal'', and ``suggest next action'' while referring to CRM records in a chat format. Available to all HubSpot users
✓ Available for all plans (limitations apply depending on functionality)
🤖
Layer 2 — Agents
Breeze Prospecting Agent
An AI agent that autonomously performs research on target contacts, generates personalized emails, and sends them. The person in charge just sets up the "Selling Profile," and the AI ​​does the rest automatically, from conducting prospect research to sending the first email. Full-scale operation begins in spring 2025
✓ Sales Hub Pro / Enterprise (Credit consumption)
🏢
Layer 2 — Agents
Breeze Closing Agent
AI agent built into CPQ's buyer portal. While the buyer is checking the estimate, AI answers questions such as ``What is the difference between this plan and a higher-end plan?'' and ``How much does it cost to add a seat?'' in real time, supporting closing.
✓ Sales Hub Pro + Commerce Hub (CPQ required)
🧠
Layer 3 — Intelligence
Breeze Intelligence (data enrichment)
Automatically fill in blank properties (industry, number of employees, technology stack, LinkedIn URL, etc.) in contact and company records from external data sources. Reduce manual entry time for sales representatives by referencing a database of over 4 billion companies and individuals
✓ Separate credit purchase required (1 enrich = 1 credit)
📊
Layer 1 — Copilot / Deal AI
AI Deal Score
Machine learning is performed on past orders and orders lost, and the probability of winning each deal currently in progress is scored on a scale of 0 to 100. The discrepancy between the subjective accuracy of the person in charge and the AI ​​score will be the most important topic in forecast review (detailed in Chapter 8)
✓ Sales Hub Pro/Enterprise (requires data of 200 or more items)
🎙️
Layer 1 — Copilot / Conversation AI
conversation intelligence
Automatic recording and transcription of calls, AI summary, and automatic detection of Deal Risk signals. Detects keywords such as “competition mentions,” “budget concerns,” and “decision maker absent” and flags deal records as risks in real time.
✓ Sales Hub Pro/Enterprise (within calling function)
💡 Essential differences between Breeze AI and ChatGPT

The biggest difference between general-purpose AI (ChatGPT etc.) and Breeze is "Accessing CRM Data"is. Breeze generates an email by referring to the following context: ``VP Nakamura of ABC Corporation visited the pricing page last week, the deal is scheduled to close on March 15th, and the past two calls have raised budget concerns.'' The core of Breeze's value in sales is that it eliminates the need to manually paste this context into general-purpose AI each time.

Section 11-2

Configuring and operating Breeze Prospecting Agent

Breeze Prospecting Agent is Sales Hub's most innovative AI-powered feature. Previously, each email required 15 to 30 minutes of work: ``Check list → Research on LinkedIn → Think about email text → Send → Track.''Prospecting Agent performs this entire process autonomously. All the rep needs to do is set up a Selling Profile and pass the target list.

Prospecting Agent operation flow

🤖
Breeze Prospecting Agent — Operation flow (per contact)
~100 credits/run
Pre-configuration (performed only once by the person in charge)
Selling Profile settings + Target list assignment
Register your company's value proposition, ideal customer image, messages to avoid, the tone of the person in charge, and available sequences in your Selling Profile. Set once and automatically applied to all AI outreach
🔍
Step 1: Context collection
Load contact/company record information
Automatically reference contact/company properties (title, industry, size, past conversation history, visited pages, past deals) in HubSpot CRM. Leverage data enriched with Breeze Intelligence
🌐
Step 2: External research
Automatically investigate LinkedIn, company sites, and news
AI autonomously collects the target contact's LinkedIn profile, recent posts, company news, recruitment information, and press releases. Automatically generate contextual information to “use recent events in emails”
✏️
Step 3: Email generation
Generate personalized emails by combining Selling Profile × CRM data × external research
Automatically generate emails that don't look like templates, combining references to the recipient's recent LinkedIn posts, implementation examples at other companies in the same industry, and issues unique to the recipient's position. Create subject line, body text, and CTA all at once
📤
Step 4: Approve & Submit
Confirmation by person in charge → Approval → Automatic transmission (or automatic transmission mode)
The default is "Send after confirmation and approval by the person in charge." Once the reliability level is high enough, you can switch to "automatic sending mode." Emails awaiting approval are listed in the Prospecting Agent tab in Sales Workspace
📊
Step 5: Track & suggest next actions
Detects opens, clicks, and replies and suggests next action
It automatically tracks openings, link clicks, and replies to sent emails, and automatically controls ``reply → stop the sequence and notify the person in charge'' and ``unopen → suggest follow-up timing.'' Persons in charge can focus only on “those who responded”
⚠️ Prospecting Agent is not a complete replacement

What is Prospecting Agent most effective for?“Tier 2/Tier 3 mass outreach”is. The approach to the most important Tier 1 accounts should be in ``human language,'' which is an AI-generated draft that has been significantly modified by the person in charge. Sentences that make the other person feel like they were written by an AI will have the opposite effect at Tier 1. It is important to clarify the division of roles between AI's "quantity and speed" and humans' "quality and relationship building."

Section 11-3

Designing the Selling Profile: Giving the “context” for AI to work correctly

The output quality of Breeze AI is Determined by the design quality of Selling Profile. Selling Profile is "a set of information that you can teach AI about your company and services." If this is ambiguous, AI will only be able to generate generic and thin emails. Information such as “We are supporting the implementation of HubSpot Sales Hub” is not enough.Customer pain, competitive advantage, NG messages, and the tone of the person in chargeBy making specific settings, the output of AI can be dramatically improved.

⚙️ Selling Profile setting example (HubSpot implementation support company Taro Yamada)
Setting location: Sales → Prospecting → Selling Profile
🏢 Product/Service Overview
Value proposition (1-3 sentences)
Increase your sales team's order win rate and forecast accuracy by supporting the implementation and utilization of HubSpot. On average, the Closed Won rate improved by 30% within 6 months of implementation.
Ideal customer image (ICP)
B2B SaaS/Services companies with 50-500 employees. 5 or more sales people. Currently managed using Excel or an old CRM, data is scattered.
Common customer challenges
Management methods vary by salesperson - Forecasting relies on intuition - Lead follow-up is overlooked - Marketing and sales work separately
🎯 Outreach settings
Sequence to use
Uses "Tier2_SaaS_Outbound_v3" (7 steps, 23 days). AI personalizes only the first email.
the tone of the person in charge
Friendly and professional. Do not use body language. Use “your company” rather than “your company.” Maximum length is 3 paragraphs.
NG message (Never write)
Don't use generic phrases such as "Our solutions are the best in the industry," "Any time is convenient for you," and "We'd love to talk to you."
✅ Decide whether the Selling Profile is “unified by team” or “individual”

Selling Profile can be set on a per-user basis. Managers can manage and update profiles if they want to use a unified profile for their team, or set them individually if they want to reflect individual differences in writing style.Value proposition, ICP, and NG messages are unified as a teamTone sequence selection is individualThis hybrid design is effective in many organizations.

Section 11-4

Limitations of leveraging and personalizing AI email generation

Breeze Copilot's "email generation" feature generates the first draft of an email in seconds based on information in contact records. It can be used as a step in a sequence or as an individual email. but“Just because it was generated by AI doesn’t mean it’s accurate”All personnel need to understand this point.

AI-generated emails — real output examples and annotations

What AI email generation can do and what it can't do

itemWhat AI is good atThings that AI is weak at and easy to make mistakes
Information gathering Highly accurate collection and summarization of public information (LinkedIn, press releases, recruitment information) It is impossible to understand the context of non-disclosed internal information and information that is only communicated orally.
Sentence generation Good at generating sentences in natural Japanese, specified tone, and specified length A ``machine-like'' feel may remain. For Tier 1 opponents, adjustments by the person in charge are required.
factual accuracy Accurately refer to data recorded in CRM (deal amount, stage, past calls) Information from external research may be outdated, exaggerated, or misinterpreted. Fact checking is required before sending
personalized We are good at personalization based on job title, industry, and recent signals. Only the person in charge can gain a deep understanding of the individual, such as ``this person's personality, preferences, and habits of speech in past conversations.''
Section 11-5

Automated Conversation Intelligence and Deal Risk Detection

HubSpot's Conversation Intelligence automatically transcribes calls recorded using Calling.AI analyzes content and displays Deal Risk signals in real-time on opportunity recordsIt is a function to do. Officially released in Fall 2025, the system will automatically visualize how the content of calls affects the prospects for business negotiations.

📝
Automatic transcription of calls
Automatically convert recorded calls to text using HubSpot's calling feature. The person in charge can concentrate on the conversation. After transcription, search, keyword extraction, and AI summary are automatically generated. You can text search past calls, so you can instantly check what was said in that business meeting.
📋
Automatic generation of AI call summaries
After a call ends, a summary of today's call, agreements, and next actions are automatically generated and automatically saved in the activity timeline of the deal record. Reduce the amount of time your reps spend taking manual notes after calls and maintain data quality in your CRM
🎯
Creating coaching clips
Detected words: “Can it work with existing systems?” “Is the security okay?” “IT department is concerned”
⚠️
→ Attend SE (technical staff) to business negotiations and provide technical specifications and security materials
AI analyzes call transcripts to automatically detect opportunity risk signals and display them on opportunity records. The system can quickly visualize "red flags" that the person in charge is unaware of, allowing them to intervene before it's too late.

6 Signals Automatically Detected as Deal Risk

⚔️
competitive comparison
Detected words: “We are also talking to 〇〇 company” “Comparing with other tools”
→ Sharing of competitive battle cards, re-presentation of differentiation points, and setting of superior call
💰
budget concerns
Detected words: "I'm on a tight budget," "I want to keep costs down," "Will the approval pass?"
→ Presentation of ROI materials, proposal of phased implementation plan, and setting up a place for CFO to be present
👤
No decision maker
Detected words: "I'll check with you," "I'll ask the manager," "I'm not the one making the decision."
→ Request Champion to act as a bridge to the decision maker / Set up a call for top players as soon as possible
📅
Postponing the timeline
Detected words: “I want to do it next term,” “It’s difficult this term,” “A little later.”
→ Identify the root cause of the postponement and re-present the reasons for moving in this period (costs, risks)
📉
Decrease in engagement
Detection: Compared to last time, meeting time has been shortened, number of questions has decreased, and replies have become slower.
→ Approach and reconfirm Champion's enthusiasm through different channels (phone, LinkedIn)
🔧
technical concerns
Detected words: “Can it work with existing systems?” “Is the security okay?” “IT department is concerned”
→ Attend SE (technical staff) to business negotiations and provide technical specifications and security materials
Section 11-6

AI credit management and ROI thinking

Some features of Breeze AI are AI creditsconsume. From September 2025, Prospecting Agent action execution and Breeze Intelligence data enrichment will be based on credit consumption. RevOps requires a design that systematically manages credits and maximizes ROI.

🤖
Prospecting Agent
1Outreach execution
about 100
Credit / Execution
Complete cycle of research, email generation, and sending. 100 outreaches per month = approximately 10,000 credits
🏢
Breeze Intelligence
Enriching company records
1
credits/records
Complementary industry, number of employees, technology stack, etc. Records once enriched are not re-consumed
👤
Breeze Intelligence
Contact enrichment
1
credits/records
Completion of email address, LinkedIn URL, and job title. When importing large quantities, estimate the required amount in advance.

AI credit ROI estimation and management policy

scenarioconsumption creditExpected effectROI judgment
Prospecting Agent Outreach to 100 Tier 2 Accounts Approximately 10,000 credits Expected response rate of 5-8% → Acquired 5-8 meetings. Reduction in man-hours for personnel: Previously 25 hours → 2 hours ◎ High ROI (man-hour reduction + meeting acquisition)
Enrich all existing contacts (5,000) at once 5,000 credits Improved scoring accuracy and eliminated blank properties. However, old data may not be updated. △ Medium (should be focused on records with high priority)
Independent execution of Prospecting Agent for Tier 1 accounts (20 companies) Approximately 2,000 credits If you send an AI-generated email as is, there is a risk that it will look like a template. If the person in charge makes extensive rework, the benefits of AI will diminish. ✗ Deprecated (Tier 1 prefers manual outreach)
⚡ Credits are prioritized and allocated based on “target × expected return”

AI credits are a finite resource. RevOps determines the monthly allocation policy,Priority allocation to “Tier 2 Mass Outreach” and “New Import Record Enrichment”We recommend that you do so. You can check the remaining credit amount from "Settings → Account and Billing → Breeze AI Credit". Prospecting Agent will stop when credits are exhausted at the end of the month, so allocate them systematically at the beginning of the month.

📌 Chapter 11 Summary

Breeze has evolved from an “auxiliary tool” to an “autonomous agent”

By using the three layers of Copilot (assistant), Agents (autonomous execution), and Intelligence (data completion), you can maximize the benefits of AI in each phase of sales activities. Rather than "leaving everything to AI," understand the strengths and weaknesses of each function and use them appropriately.

The quality of a Prospecting Agent is determined by its Selling Profile

By specifically setting the value proposition, ICP, NG message, and tone of the person in charge, the output quality of AI can be dramatically improved. Information such as "We are supporting the implementation of HubSpot" will only generate a generic email. It's worth the time for initial setup.

AI is responsible for “quantity” in Tiers 2 and 3, and humans are responsible for “quality” in Tier 1.

Prospecting Agent shows its true value in mass outreach to Tier 2 and 3. When approaching the most important Tier 1 accounts, it is important that the person in charge makes significant adjustments based on the AI-generated draft and delivers it in ``human words.''

Deal Risk makes ``unnoticed problems'' visible at an early stage

Conversation intelligence detects six risk signals (competition, budget, decision maker, timeline, engagement, and technology) that reveal issues that people in charge may not be aware of. The golden rule is to take immediate action when you see a signal.

AI credits will be prioritized and allocated based on ROI estimates.

Understand the costs of Prospecting Agent (approximately 100 credits/execution) and Breeze Intelligence (1 credit/record) and prioritize allocation to Tier 2 mass outreach and new record enrichment. A solo run to Tier 1 is not cost-effective.

AI does not replace “judgment”; it replaces “work”

AI can automate repetitive tasks such as research, drafting, data entry, and tracking. Only humans can continue to be able to ``build a relationship of trust with this person,'' ``identify the real blocker in this business negotiation,'' and ``decide on the final move in the negotiation.'' Investing the time freed up by AI in "judgment and relationship building" will become the essential value for salespeople in the AI ​​era.

Next Chapter
Chapter 12: Operational Design — Build a sustainable Sales Hub operational structure →