HubSpot Breeze: Components, Credits, Data Requirements and Governance

HubSpot Breeze is budgeted as though it were one product with one price, and three surprises follow. A smart property is filled across 5,000 company records, and HubSpot charges 10 credits for every record whether or not a value comes back. Summaries on contact records stop reflecting new calls late in the month while the credit balance is untouched, because conversation analysis draws on a separate capacity pooled from paid seats. A workflow that sorts inbound leads by an AI-generated category starts sending them down its no-match branch, because once credits run out the AI action fails and writes an empty value.

This article explains what HubSpot Breeze covers in September 2026 and how each part is paid for. It starts with the names HubSpot now gives the assistant, the agents and the enrichment features, then sets out which features consume credits and what stops each at its limit. It prices a sample month, follows what each feature reads from the CRM, and sets out the governance to settle before agents go live, closing with a procedure, costs and returns, symptoms, common questions and the limits of the evidence.

Breeze is HubSpot's name for its AI, described in the navigation of the Product and Services Catalog as "AI agents and features that power the entire platform". A HubSpot Credit is the unit in which usage-based features are billed, reset monthly with no rollover (Manage HubSpot Credits). Conversational context is HubSpot's background analysis of logged call transcripts, emails, meeting transcripts and chats, which feeds AI summaries and is capped by the number of paid seats (communication analysis limits).

Breeze AI Components and Their Current Names

No HubSpot page found maps the Breeze names to the current ones, so the table below is assembled from evidence observed on 27 September 2026. The Breeze Assistant product page states that Agent Hub was formerly Breeze Agents and the agent builder formerly Breeze studio. The Agent Hub product page gives "Breeze" as its alternate name in the page's structured data. Old addresses redirect: www.hubspot.com/breeze to that page, the Breeze Copilot product address to the Breeze Assistant page, and the knowledge base article on Breeze Intelligence to Get started with data enrichment.

The name therefore identifies nothing a business can buy or price as one item. The finding of this article follows: Breeze features draw on three separate meters, namely none at all, HubSpot Credits, and a conversational context capacity set by seat count, and each meter stops work in a different way at its limit.

The agents, eleven in the catalog and eight labelled Beta, are treated in HubSpot Agent Hub, and appear here only where they share a meter, a setting or a data dependency with the rest of Breeze.

Breeze Assistant Scope and Defaults

Breeze Assistant is HubSpot's conversational assistant, opened from the top navigation bar, and its product page states that it is included with every subscription at no additional cost. The AI adoption guide recommends giving users access because it works without using HubSpot Credits. Use Breeze Assistant caps content generation at 100 requests per account in any rolling minute, and creating an agent or workflow through the assistant requires Professional or Enterprise.

Two defaults matter for governance: access is on for every user, and conversation history is on; the assistant can also store memories of each user's job title, tone and recurring instructions. Artifacts, the HTML pages, documents and presentations it produces, can be published at a read-only public URL with an optional password, and the article names no permission that governs public sharing. Projects, which replaced custom assistants on 13 July 2026, can hold a Write to the CRM tool that creates or updates records during a chat.

User permissions determine which actions the assistant can perform, and its context panel lists the records, files, connected apps and memories behind each response.

Breeze Intelligence as Data Enrichment

What the knowledge base once addressed as Breeze Intelligence now appears as data enrichment, available with Starter, Professional or Enterprise editions of Marketing, Sales, Service, Data and Content Hub. HubSpot matches a contact on first name, last name and work email, and a company on its domain (How data enrichment works). The getting-started article states that enrichment does not consume HubSpot Credits.

Automatic enrichment is off until a Super Admin turns it on, fills only empty properties, and leaves values set by a user or another system alone; continuous enrichment refreshes enriched values monthly and stops on a property once a user or system edits it (enrichment settings). Two routes ignore the Mapping tab's overwrite rules: the Enrich record workflow action set to overwrite, and conversational enrichment, a public beta that reads one-to-one sales emails, updates contact properties and can create contacts from threads, free of credits while in beta.

Three neighbouring features are billed differently. Form shortening hides fields enrichment can fill, works only in the updated form editor, and has used no credits since 3 September 2025. Smart properties, built with the data agent, answer a prompt per record from web research, the company website, another property or logged activities, and consume credits even when no value is filled. Buyer intent is free to browse, while adding an identified company costs credits once and tracking one costs them monthly (buyer intent questions).

HubSpot Breeze Credits, Allowances and Exhaustion

Included credits follow the highest tier held, not the number of hubs. The catalog grants 500, 3,000 or 5,000 a month at Starter, Professional or Enterprise for Smart CRM, Marketing, Sales, Service, Content or Revenue Hub, and 500, 5,000 or 10,000 with Data Hub or the Customer Platform bundle. Only users on paid seats or Partner Seats can use credits, and extra credits cost $0.010 each as overage or $10 for each monthly pack of 1,000.

Outside the agents, the rate sheet charges 10 credits per AI action in a workflow, 10 per company tracked for intent each month, 50 per company per month for a custom signal (Beta), and 10 per prompt per record for a smart property. The credits article warns that beta features may begin consuming credits, with thirty days' notice an aim rather than a promise.

An account that has bought no credits pauses usage-based features until the reset date; one that has bought credits pays overage or is upgraded for the rest of its term, depending on its billing setting. Individual features fail in their own ways: the Data Agent: Custom prompt workflow action writes null outputs, while buyer intent blocks adding and tracking and keeps existing intent data visible.

Conversational context is a separate meter that credits do not touch, used by the CRM cards that summarise contacts and companies and by brand identity. The catalog allows 50, 250 or 350 records per paid seat per month at Starter, Professional or Enterprise, pooled across the account. At the limit, analysis of new communications stops until the reset or a $50 pack of 5,000 more records, while logging continues.

Breeze Consumption on a Sample Professional Account

The figures below are sample data built from catalog rates read on 27 September 2026; they describe no client and are not a quote. The sample account holds Sales Hub and Marketing Hub Professional with 14 paid seats, so it receives 3,000 credits a month rather than 6,000, and conversational context capacity of 14 × 250 = 3,500 records.

With overage billing, the month bills 4,700 credits beyond the included 3,000, which is $47. With nothing purchased, the outcome depends on when charges land, which varies by account. If the two intent charges come first, they take 2,200 credits, and the remaining 800 pay for 80 of the 550 per-record prompts. The other 470 leave a smart property unfilled or write a null output that no workflow branch matches.

Conversational context runs out on its own schedule. Ten sellers log 330 emails and transcripts each, four Core Seat users log 50 each, and the shared inbox takes 700 chats: 4,200 records, or 140 a day over 30 days. The 75 per cent notice arrives on day 19 (2,625 ÷ 140 = 18.75), capacity runs out at the end of day 25 (3,500 ÷ 140), and the 700 communications of the last five days are stored but not analysed.

The largest avoidable line is a one-time run. Filling the same web-research smart property on all 5,000 existing companies costs 50,000 credits, or $500, and 850 of those companies carry no usable domain. Restricting the run to the 4,150 with a valid domain saves 8,500 credits, or $85, since the charge applies whether or not a value is filled.

CRM Data Requirements for Breeze Output

Each Breeze feature reads named fields, and those fields cap what it returns. Enrichment needs a business email on a contact and a domain on a company; addresses from Gmail, Yahoo and other free services are excluded, and a malformed domain such as hubspot..com is ineligible (data test). Web-research smart properties key on the company domain. The Data Agent: Custom prompt workflow action sees only the properties inserted into its prompt, runs on a model with no internet connection, and takes a prompt of up to 3,000 characters (AI actions in workflows).

In the sample, 700 of 5,000 companies have no domain and 150 a malformed one, so enrichment can reach no more than 4,150 companies, or 83 per cent. Of 15,000 contacts, 3,300 use personal addresses and 450 have no email, leaving 11,250, or 75 per cent. These are ceilings, since the match rate also depends on whether HubSpot holds data for each domain.

Duplicates degrade output too. The adoption guide warns that duplicates can lead HubSpot's AI to outdated or conflicting information, and companies created through the API, including by sync apps, are not deduplicated by domain (deduplication). The catalog lists duplicate management only at Professional and Enterprise, so Starter accounts lack the tool that lists duplicates for review.

Conversational context depends on logging in the same way: an unrecorded call or an unlogged email never enters the analysis. The data layer beneath every hub, and what AI features reach through its associations, is the subject of HubSpot Smart CRM. For Breeze an empty field is either unreachable or billed without return, so property completeness per segment, measured as in CRM adoption measurement, comes before any per-record run.

Governance Before HubSpot Breeze Agents Go Live

Several governance decisions are made by default. In AI settings, which only a Super Admin can change, generative AI access, CRM data, customer conversation data and Breeze Assistant access are on by default, Files data is off, and AI model training is on. Accounts with Sensitive Data are opted out of training and cannot opt in; for every other account, training stays on until someone turns it off, and each change is recorded in the audit log.

The adoption guide recommends limiting agents and assistants to specific users, and marks the permissions that spend credits: Agent builder, Customer agent editor, both prospecting agent permissions, Data Studio use, Data enrichment access and Data Agent access. Agents and projects carry their own access settings, each a decision to record.

Credit governance works at three levels. The account maximum, set 50 per cent above total monthly credits once credits are purchased, pauses every credit feature; a feature-level limit caps one tool without reserving credits for it; and action-level limits exist only for buyer intent. Content governance covers public artifacts and the Write to the CRM tool, and data governance covers the overwrite rules and the conversational enrichment that bypasses them.

Breeze Readiness Procedure

1. List each intended Breeze feature with its meter, billing unit and expected monthly count, and any one-time run with its record count.
2. In Settings > Account management > AI, record a decision on each Access switch, including Files data, Breeze Assistant access and AI model training.
3. In Data Management > Data Enrichment, click Scan for enrichment gaps and record the match rate for contacts and companies.
4. Fill Company domain name, correct malformed domains, review personal addresses, and run duplicate management over each target segment.
5. In the enrichment settings, set an overwrite rule per property on the Mapping tab, decide on automatic and continuous enrichment, and review the Conversational Enrichment tab.
6. In Users & Teams, grant Data enrichment access, Data Agent access and agent permissions to named users only.
7. In Account & Billing, on the HubSpot Credits tab, set maximum monthly credits, a feature credit limit per credit-consuming tool, and action-level limits for intent.
8. Multiply paid seats by the per-seat conversational context allowance and compare the result with last month's logged emails, transcripts and chats.
9. Verify after the first month: compare the count of records with Has been enriched set to True against the step 3 match rate, set filled smart property values against the data agent credits shown under Manage usage by feature, at 10 a prompt, test the workflow AI action on three records for a non-null output, and note the day any conversational context notice arrived.

Costs, Returns and Selection Conditions for Breeze

Breeze Assistant, data enrichment and form shortening consume no credits, so an account can take a first return before any purchase: enriched fields, shorter forms and summaries on records. The metered parts cost one cent a credit, so a smart property answer or workflow classification costs $0.10 a record, as does a month of intent tracking for one company.

The credit prices do not decide the case: the sample overage cost $47 and the unfiltered backfill wasted $85, while the cost nobody invoices is a workflow writing null values for the leads it was built to route. Preparation is the cost the rate sheet omits: domains and business emails repaired, duplicates merged, overwrite rules set and permissions limited.

The Breeze Assistant page reports averages such as 2.7 times more deals closed with the assistant, without saying how users were selected or compared, so the figures are vendor claims rather than expected returns. The case is strongest where companies carry clean domains and contacts use business email, the population enrichment and web research can reach. It is weakest where contacts arrive with personal addresses, which enrichment excludes by design, and where per-record runs exceed the allowance with no limit set.

Two alternatives win on specific axes. A required form field or import column wins on provenance, because the value comes from the customer or source system at no charge, and it reaches contacts with personal addresses. A saved report wins on reproducibility, returning the same result for the same filters, whereas an assistant answer depends on the prompt, the retrieved context and stored memories. The firm is a HubSpot Solutions Partner and is paid for the data preparation described here, which is the interest to weigh against this section.

Breeze Symptoms and Their Mechanisms

A workflow branch that depends on an AI-generated value sends records to its no-match path on some days. The custom prompt action fails once credits run out and writes null outputs; match the feature credit limit to the workflow's volume, or add a branch for the empty value.

A smart property shows blanks although credits were consumed. Filling bills per record whether or not a value returns, and web research has no domain to search on for some companies; fill from a segment filtered to known domains.

Enrichment leaves an outdated job title unchanged. Automatic enrichment fills only empty properties, and continuous enrichment stops on a property once someone edits it; change that property's overwrite rule, or enrich manually with overwrite selected.

Protected contact properties change, and new contacts appear from email threads. Conversational enrichment updates and creates contacts without following the Mapping tab, so its switches on the Conversational Enrichment tab need to be off.

Summaries on record cards stop reflecting recent calls late in the month while credits remain. The account has reached its conversational context capacity, which pauses new analysis until the reset; the 75 per cent notice is the point to buy an increase or accept the gap.

Buyer intent refuses to track more companies after others were untracked. Untracking frees no capacity within the current period, so the remedy is a higher action-level limit or the next reset.

Frequently Asked Questions

What is HubSpot Breeze: a product, a feature set or a name?

A name for HubSpot's AI, surviving in Breeze Assistant, while the agents sit in Agent Hub and the former Breeze Intelligence features under data enrichment.

Does HubSpot Breeze pricing include a separate licence fee?

No. Breeze Assistant and enrichment are included, and metered features draw credits at $0.010 each, first from the 500, 3,000 or 5,000 included monthly at the highest tier held.

Is HubSpot Breeze Copilot the same tool as Breeze Assistant?

The Breeze Copilot address on hubspot.com redirects to the Breeze Assistant page, the only evidence of the rename found, and current pages use Breeze Assistant alone.

Does HubSpot Breeze Intelligence data enrichment consume credits?

Contact and company enrichment does not, nor has form shortening since 3 September 2025. Smart properties cost 10 credits per record prompt, and buyer intent charges for adding and tracking companies.

Which HubSpot Breeze credits come with a Starter subscription?

Five hundred a month, whichever Starter product is held: enough for 50 smart property answers or 50 companies tracked for intent. Workflow AI actions need Professional or Enterprise, the step set out in HubSpot Starter vs Professional.

Are HubSpot Breeze agents still offered under the Breeze name?

No. The Breeze Assistant page states that Agent Hub was formerly Breeze Agents, and agents are activated, built and monitored there.

Boundaries of the Evidence, September 2026

Every product fact here is as documented in September 2026, read from hubspot.com, the knowledge base and the catalog on 27 September 2026, which are authoritative over this text. The agents themselves, AEO, HubSpot Work and Data Studio are out of scope. Conversational enrichment, custom signals, custom agents and conversation summaries are in beta.

The evidence has four limits. The mapping of old names rests on two "formerly" statements, redirects and structured data, because no HubSpot statement of the renaming was found. The sources disagree twice: the knowledge base lists enrichment for Smart CRM at Professional and Enterprise while the catalog adds Starter, and the adoption guide marks Data enrichment access as spending credits while the enrichment article says enrichment does not. HubSpot does not define how a thread counts against conversational context, so the sample counts each email, transcript or chat as one record. The only outcome figures found are vendor averages without a method, so the sample prices Breeze without showing that it pays.

In Summary

HubSpot Breeze in September 2026 is a name for HubSpot's AI, not a product with a price. Breeze Assistant keeps the name and is included with every subscription; Breeze Agents and Breeze studio became Agent Hub and the agent builder; and the features once addressed as Breeze Intelligence sit under data enrichment, beside smart properties and buyer intent.

Those parts draw on three meters. The assistant, enrichment and form shortening use none. Smart properties, workflow AI actions, intent tracking and agents draw on 500, 3,000 or 5,000 credits a month set by the highest tier held, and a workflow action that runs out writes a null value. Summaries built from calls and emails draw on 50, 250 or 350 records per paid seat a month, whatever the credit balance.

Two counts belong before any per-record run, companies with a valid domain and contacts with a business email, since enrichment cannot reach the rest and a web-research smart property bills for them anyway. On the sample account those shares were 83 and 75 per cent, and the month's whole credit overage, $47, cost less than the $85 one unfiltered backfill spent on companies with no domain.