Start with Breeze Assistant, which comes with every HubSpot plan, then add one agent that matches your busiest queue: Customer Agent for support chats (Professional or Enterprise, 50 credits, about $0.50, per resolved conversation) or Prospecting Agent for outbound leads (100 credits, about $1, per lead). Leave the rest off until you've watched a month of credit use.
The money works differently from a normal subscription. Breeze agents run on HubSpot Credits at $10 per 1,000, Starter includes 500 credits a month and Professional 3,000, and unused credits don't roll over. HubSpot's help pages add that if you use up your included credits without buying more, credit-based features simply pause until the next monthly reset. For a small team, the AI budget is really a monthly credit budget.
Breeze in plain terms: an assistant, some agents and a meter
HubSpot's AI has four parts, and it helps to keep them separate:
- Breeze Assistant is the chat panel inside HubSpot. HubSpot lists it for all products and plans. It drafts content, summarises records, prepares you for meetings and completes tasks, and it can save prompts, keep memories and group chats into projects.
- Agent Hub is the workspace where agents are switched on, configured and monitored. HubSpot's help pages list it from Starter upwards, while noting that some HubSpot-built agents need Professional or Enterprise, and that agentic workflows do too.
- HubSpot-built agents do specific jobs: Customer Agent answers support conversations, Prospecting Agent researches companies and drafts outreach, Data Agent enriches and maintains CRM data, and Deal Progression suggests next steps on deals. HubSpot's Agent Hub page also lists content, campaign, nurture and revenue agents; check which ones your plan includes before planning around them.
- HubSpot Credits are the meter. Agents and many AI features draw on them, and custom agents you build yourself in the agent builder consume them too.
One piece of history matters if you've used HubSpot for a while. On 23 July 2026 HubSpot sunset five Breeze agents: Social Post, ABM Landing Page, RFP, Cross-sell/Upsell and Sales to Marketing Feedback. Existing installs keep working, but they can't be newly installed, so don't build a process around one now.
What each plan gives you
This summary follows HubSpot's knowledge base as updated in September 2026. HubSpot changes packaging often and its pricing page tends to show promotions, so treat it as a checklist to confirm against your own account.
| Plan | AI you can use | Credits included | What that buys |
|---|---|---|---|
| Free tools | Breeze Assistant, with some features reserved for paid plans | No monthly allowance stated | Drafting and summaries |
| Starter (list price $20 per seat a month) | Breeze Assistant; Agent Hub; Prospecting Agent appears in HubSpot's Starter documentation, running on credits | 500 a month | At most 5 prospecting leads a month before you buy more |
| Professional | All of the above, plus Customer Agent and agentic workflows | 3,000 a month | For example, 60 resolved Customer Agent conversations |
| Enterprise | As Professional, with more controls | 5,000 a month | For example, 100 resolved conversations |
A common mistake shows why the table matters. An illustrative locksmith on Starter spent an afternoon looking for Customer Agent to answer out-of-hours chats, before finding that it needs a Professional or Enterprise subscription. HubSpot often runs steep first-year discounts, so compare list prices, not the promotion, if an upgrade is what the agent would take. Whether that upgrade pays for itself is weighed up in whether HubSpot is worth paying for just for its AI features.
Switch on first: Breeze Assistant for everyday CRM work
Breeze Assistant uses no special setup beyond turning on user access in your AI settings, and it's where most small teams get their first real time savings. Four jobs worth trying in the first week:
- Before a call: "Summarise this contact's history: last three emails, open deals, any complaints, and what we promised last time."
- After a call: paste your notes and ask for a follow-up email and three tasks with due dates.
- Before a meeting: with your calendar connected, ask it to prepare you for tomorrow's meetings.
- For the team: ask which deals haven't been touched in 14 days, and draft a nudge for each.
Here's the first job for an illustrative roofing contractor, with a typical reply:
Prompt: Summarise [contact]'s history with us: last 3 emails,
open deals, complaints, and anything we promised.
Reply (illustrative):
- Quote sent 12 Aug for rear roof repair; no reply since
a follow-up on 19 Aug.
- Open deal: "Rear roof repair", stage: quote sent.
- No complaints logged.
- Promise: "we'll hold the price until end of September"
(email of 12 Aug).That last line is the valuable one, because it's easy to forget and awkward to break. What to check: the assistant only knows what's logged in HubSpot. If the promise was made by phone and never noted, it won't appear, so the summary is only as good as your record-keeping. That's the strongest reason to tidy the CRM before adding agents, and cleaning up a messy CRM with AI covers how. HubSpot also caps content generation at 100 requests per account in any rolling minute, which a small team won't notice but a runaway workflow might.
Customer Agent: when 50 credits a resolution pays
Customer Agent answers questions on your support channels using content you give it: your website pages, HubSpot knowledge articles and similar sources. Based on how confident it is, it answers with a cited source, asks a clarifying question or hands the conversation to a person. It needs a Professional or Enterprise subscription, HubSpot Credits to deploy, and a connected channel such as live chat. Credits are used only when it delivers a resolution, and HubSpot offers 14 days of free access the first time you set it up.
Here is the arithmetic for an illustrative cleaning company with 14 staff on a Professional plan, handling about 600 chats a month about prices for regular cleans, what's included, rescheduling and key collection:
- During the free trial, the agent resolved about 40% of conversations. At full volume that's roughly 240 resolutions a month.
- 240 resolutions at 50 credits each is 12,000 credits.
- The plan includes 3,000 credits, shared with any other AI features, so about 9,000 extra credits are needed: at $10 per 1,000, around $90 a month.
- If each of those chats would have taken a staff member five minutes, that's about 20 hours a month handled by the agent.
At roughly $90 for 20 hours, the case is clear for this firm, as long as the answers are right. Two details from HubSpot's documentation shaped the rollout. First, previewing and testing, including batch testing with generated questions, doesn't use credits, so they ran every common question through it before going live. Second, once a customer agent is created it can't be deleted, only edited or paused, so they named and configured it as though it would be permanent.
Before relying on the 40% figure, check how HubSpot counts a resolution in your usage log. With any per-resolution billing, a customer who simply gives up and closes the chat can look like success. The firm also set hard handoff rules for complaints, anything about keys or alarm codes, and commercial contract pricing, because those are the answers where a mistake costs more than a chat. Your business answers for what the agent says, which is set out in who is liable when your AI chatbot gets it wrong.
A realistic set-up mistake is worth avoiding. An illustrative removals firm synced its whole website as the agent's knowledge, including an old price page that was still live but no longer linked from the menu. The agent quoted last year's prices for a week. HubSpot's website sync can crawl up to 5,000 pages on a domain, and it lets you include or exclude paths, so point it only at current service and FAQ pages and delete stale pages from the site rather than trusting that nobody finds them.
Prospecting Agent: 100 credits a lead adds up quickly
Prospecting Agent watches for buying signals, such as a company announcing funding, a new product or research into topics you care about, then identifies contacts and drafts personalised outreach. You guide it with "plays": instructions describing the kind of company, the people to contact and your pitch. It uses credits each time it generates outreach for a lead, at 100 credits per lead.
For an illustrative HVAC installer chasing commercial maintenance contracts, the sums look like this. Fifty recommended leads a month is 5,000 credits, or about $50 at list. On Professional, the 3,000 included credits would be gone by lead 30 if nothing else used them. On Starter, 500 credits cover five leads. The money is modest; the real cost is review time, because every draft goes out under your name and needs checking by someone who knows the trade.
There's also a data decision. HubSpot's requirements for the agent include switching on AI access to CRM data, customer conversation data and files data in your AI settings. That's reasonable for a sales tool, but it's a conscious choice about what HubSpot's AI can read, so make it deliberately rather than clicking through. If your outbound effort is a handful of warm contacts a month, a good Breeze Assistant prompt and your own judgement may do the job without the agent.
Agents a small team can leave off for now
Not every agent earns its credits at small scale. Reasonable candidates to postpone:
- Data Agent, unless your CRM is genuinely messy. It enriches and maintains records, which pays off with thousands of contacts and less so with a few hundred you know by name.
- Deal Progression, if your pipeline has a dozen open deals that you review every Monday anyway.
- AEO, HubSpot's tool for visibility in AI-powered search, which needs Marketing Hub Professional or Enterprise.
- Custom agents built in the agent builder, until you've run a HubSpot-built agent for a month and understand your credit use. Custom agents consume credits too.
- Any of the five sunset agents, even if an old install still works, because they won't be available to reinstall.
Keeping the credit bill predictable
HubSpot's billing rules for credits changed in September 2026, and the defaults matter:
- With only included credits, credit-based features pause when you run out. Nothing is charged, but the agent stops answering.
- Once you buy extra capacity packs, going over triggers either automatic upgrades or pay-as-you-go overages. For packs bought on or after 16 September 2026, pay-as-you-go is the default; before that, auto-upgrade was, and an auto-upgrade raises your tier for the rest of your commitment term.
- After a pack purchase, HubSpot sets an account-level spend limit 50% above your monthly credits, and emails Super Admins and Billing Admins as usage climbs.
- You can set a maximum monthly credit limit for the account and for individual tools. Use it.
A one-page monthly credit plan keeps everyone honest. Here is one for an illustrative landscaping company on Professional, filled in:
HUBSPOT CREDIT PLAN Month: [month]
Included credits (Professional): 3,000
Extra pack purchased: 5,000
Total available: 8,000
Tool Expected use Tool limit set
Customer Agent 90 x 50 = 4,500 5,000
Prospecting Agent 20 x 100 = 2,000 2,500
Other AI features about 500 500
Buffer (remaining)
Review on the 20th: usage log checked by [first name]
If Customer Agent passes 4,000 by the 20th: check for
repeat questions and add an FAQ page instead.The last line is the useful habit. When one question drives a large share of resolutions, a clear page on your site may answer it before anyone opens a chat, which saves credits and makes the agent's job easier.
A four-week rollout that avoids surprises
Switching everything on in one afternoon is how teams end up with an agent quoting old prices and a credit alert by the 10th. A slower order works better, and it fits comfortably around normal work:
- Week 1: the assistant only. Turn on Breeze Assistant for the people who live in HubSpot, agree three prompts everyone uses (call prep, follow-up email, weekly stale-deals check) and save them. Note which answers were wrong because a record was incomplete; that's your clean-up list.
- Week 2: build and test the agent, without customers. Set up Customer Agent inside its free access period, point it at current pages only, write the handoff topics, and batch-test it with your 30 most common questions. Testing costs no credits, so be generous.
- Week 3: go live with a ceiling. Deploy to one channel, usually website chat, with a tool credit limit set at roughly what you expect to use. Read 20 conversations every couple of days.
- Week 4: decide with numbers. Compare resolutions, handoffs and credits used against the hours your team got back. Keep it, adjust it, or pause it; pausing is always available even though deleting isn't.
Only after that month does it make sense to look at Prospecting Agent or anything else. The handoff rules deserve particular care, because they decide what the agent does when it's out of its depth; when an AI chatbot should hand over to a human goes through the triggers worth setting.
Settings to check before any agent talks to customers
Before switching Customer Agent or any outward-facing agent on, run down this list:
- AI settings: which data sources HubSpot's AI may use, and which users can use AI features at all.
- Knowledge sources: only current pages and articles, with old paths excluded.
- Personality and brand voice: HubSpot offers friendly, professional, casual, empathetic and witty styles, or your configured brand voice. "Witty" is rarely right for a complaint.
- Guidelines and handoff: topics the agent must pass to a person, and what it says while doing so.
- Disclosure: make it obvious customers are talking to an AI, with a route to a person.
- Tool credit limit: set before launch, not after the first surprise.
Done in that order, Breeze gives a small team real help for a predictable monthly cost: the assistant from day one, one agent where the queue is heaviest, and everything else switched on only when the numbers say it's worth it.
Further reads
- HubSpot vs Pipedrive for AI Sales Automation in a Small Business — How HubSpot's AI compares with a leaner sales CRM.
- What Is an AI CRM and Which Features Actually Save Time? — Which AI CRM features save time in any system, not just HubSpot.
- How to Hire a HubSpot or CRM Consultant to Set Up AI Features — When it's worth paying someone to configure agents properly.
- AI Call Summaries: Log Every Phone Enquiry in Your CRM — Logging phone enquiries in the CRM so agents have context.
- How to Test a Customer Chatbot Before It Goes Live — A test plan to run during Customer Agent's free trial.
- How to Automate Sales Follow-Ups With AI Without Being Pushy — Follow-ups that feel personal, with or without Prospecting Agent.
- How to Prepare Your Small Business for AI Agents — Six things to have ready before an AI agent acts for you, with an approval table, a photography studio's four-week prep and a checklist.
- What Are AI Marketing Agents and Should a Small Business Use One? — What separates a marketing agent from a chatbot or automation, the agents small businesses can switch on now, what they cost and the guardrails to set.
- Per-Seat vs Usage-Based AI Pricing: Which Costs Less for You? — How seat, usage and hybrid AI pricing actually charge, the break-even sum to compare them, and worked examples using real 2026 prices.
- How to Pilot Your First AI Agent Without Risking Customers — Run your first AI agent in shadow mode, then behind approvals, then live in a narrow window. A café's catering agent shows each stage, cost and stop rule.
- What Is an AI Agent? A Plain-English Guide for Business Owners — What an AI agent actually is, one task followed step by step, the agents small businesses can switch on today with costs, and how to give one its first job.
- AI Agent Use Cases for Small Businesses: 10 Worked Examples — Ten AI agent jobs in six small businesses, each with the before and after, the tool, the monthly cost at list prices and what a person still has to check.
- Is It Worth Switching CRM Just for Better AI Features? — Where CRM AI features sit on each price list, the migration costs switching quotes leave out, and a three-year worked sum for a small sales team.
- Outcome-Based AI Pricing: Paying per Resolution, Task or Result — How per-resolution and per-task AI pricing works, why each vendor's definition of resolved changes the bill, and how to check what you're paying for.
- What Does an AI SDR Cost, and What Does It Actually Do? — Published AI SDR prices from about $1 a lead to $3,750 a month, the mailbox and staff costs around them, and how to work out cost per booked meeting.
- AI Features Already in Your Software: What to Switch On First — Thirteen built-in AI features in Microsoft 365, Google Workspace, Xero, QuickBooks, Canva, HubSpot and Shopify, ranked by when to switch each one on.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: HubSpot knowledge base (Understand Agent Hub; Set up the customer agent; Set up and use the prospecting agent; Use Breeze Assistant; Understand HubSpot Credits and billing, all updated September 2026); HubSpot Agent Hub product page; HubSpot pricing and credit facts checked 27 September 2026.