A CRM database degrades even when nobody touches it. Contacts change jobs, companies merge, emails become invalid, company registration numbers stay empty. Reference studies put this erosion at around 2.1% per month, meaning 22.5% of the database becomes outdated every year (MarketingSherpa research, cited in HubSpot's Database Decay simulation). Meanwhile, your teams segment, score and route leads based on wrong or empty fields.
Automatic enrichment promises to reverse the trend. With Breeze Intelligence, HubSpot natively includes a building block that completes contact and company records, detects buying intent signals and shortens forms. The temptation to activate it on day one is strong.
We think this is a mistake. Enriching an ungoverned database means paying to reproduce your own errors at scale. This article details what Breeze Intelligence actually does, why the order of the work matters (governance first, enrichment second, AI agents last) and how to structure the approach, with use cases by team.
The degradation of a CRM database is not an accident. It is a structural phenomenon, driven by four mechanisms that operate continuously.
The first is natural obsolescence. A B2B contact changes role, company or email address at a pace no team can follow manually. The field was correct when it was entered. It no longer is.
The second is inconsistent data entry. Three sales reps create the same company under three different spellings (“Société Générale”, “SG”, “Societe Generale Paris”). Each fills in the fields they consider useful, ignores the others, and sometimes invents values to get past a mandatory pipeline stage. Multiply by five years of history and two tool migrations: you get the typical portal we audit at the start of a project.
The third mechanism is uncontrolled imports. Trade show lists, exports from an old tool, purchased files, partial synchronizations with an ERP or a prospecting tool: each import can add duplicates and incompatible field formats if nobody has defined mapping rules beforehand.
The last one is the most insidious: the lack of an owner. In many organizations, nobody is responsible for CRM data quality. Marketing blames sales, sales blames the tool. Nobody fixes anything. The direct consequence: reports become wrong, teams stop trusting them, then stop filling in the tool, which further worsens quality. This loop (configuration accuracy, correct reporting, team trust, adoption) works in both directions. When it runs in reverse, no enrichment tool will fix it on its own.
Breeze Intelligence is the data building block of Breeze, HubSpot's artificial intelligence layer. It relies on a database of company and buyer profiles (notably from HubSpot's acquisition of Clearbit) to automatically complete the records in your CRM. Three capabilities structure the offer.
From an email or a domain, Breeze Intelligence completes a record's properties: firmographic data for companies (industry, headcount, estimated revenue, location, technologies used), demographic data for contacts (job title, seniority). Enrichment can be done on demand on a record, or automatically when new records are created, depending on the portal's configuration. The exact scope of enriched properties changes regularly: check HubSpot's documentation before scoping a project.
The buyer intent module identifies the companies visiting your website and cross-references this signal with the Breeze database to reveal accounts in active research, even when no form has been filled in. You define an ideal customer profile (size, industry, geography) and high-intent pages (pricing, demo, comparisons); the tool surfaces the matching accounts. It is a signal, not a lead: the company is identified, not the individual.
Form shortening uses the data Breeze already knows: when a visitor can be matched to an enriched profile, the form hides the questions whose answers are already available and only shows the fields that are actually missing. Less friction at conversion, with no loss of information for downstream teams.
The billing model changed in spring 2026, and in the right direction for getting started. Standard contact and company enrichment is now included with Core seats: improving the quality of your database no longer costs credits. What is still charged by usage, through HubSpot's shared credit pool: buyer intent tracking (around 10 credits per monitored company), advanced enrichment beyond standard fields (smart properties), Data Agent research and AI actions in workflows (around 10 credits per execution). Each subscription includes a monthly volume of credits (500 in Starter, 3,000 in Pro, 5,000 in Enterprise), extendable with packs.
Two management habits follow from this. First, activating standard enrichment is no longer up for debate: it is included, so take advantage of it right away. Second, budget advanced uses based on actual volume (number of accounts to monitor for intent, critical AI workflows) rather than discovering consumption at the end of the month. The orders of magnitude above date from the Spring Spotlight 2026: check the current HubSpot pricing , as this model is still changing.
A point of attention for European portals: data coverage varies by geographic area, and processing enriched contact data raises compliance questions that deserve serious examination.
Here is the scenario we see again and again. A team activates enrichment on a portal that contains 30% unmerged duplicates, redundant properties created over the years and no normalization rules. Breeze does its job: it completes records. All records. Including the three versions of the same account.
The result is predictable. Each duplicate becomes an enriched duplicate, therefore more credible, therefore harder to detect. Credits are wasted on records that should never have existed. Scoring and routing workflows, which now read filled-in fields, trigger actions on data that is inconsistent across records: the sales rep receives the same account twice with two different headcounts. Their trust in the tool, already fragile, does not recover.
There is a second, more subtle trap: the conflict between enriched data and declared data. If a prospect stated a headcount of 80 people in a form and Breeze suggests 200, which one is authoritative? Without a priority rule defined per property (declared data, enriched data, data entered by a sales rep), enrichment overwrites or coexists depending on how the portal happens to be configured. This is not a tool problem. It is a governance problem.
Hence the position we defend at Markentive, which structures our approach to AI in HubSpot: governance first, enrichment second, agents last. Connecting AI agents (prospecting, support, content) to a database that is neither governed nor enriched amounts to automating the production of errors at machine speed. The order of the work is not negotiable. An organization that activates Breeze agents before dealing with its duplicates is simply preparing excuses for the day the pipeline turns out to be wrong.
Three steps, in this order. The temptation to run them in parallel is understandable and almost always counterproductive: the governance rules come from the audit, and the enrichment configuration comes from the rules.
Before any configuration, map the portal's data assets: number of records per object, completeness rate of critical properties, estimated volume of duplicates (contacts and companies), redundant or obsolete properties, data sources feeding the database (forms, imports, synchronizations, manual entry). The audit must also qualify uses: which properties actually feed segmentations, scoring, workflows or reports? A property nobody reads does not need to be enriched.
For data migrations, we apply a four-stage protocol: basic portal configuration, data migration and cleaning, advanced configuration and imports, then systematic verification before go-live. This sequence, proven on migrations with extensive history, applies as is to an enrichment project: Breeze comes in at the third stage, never the first.
This is the core of the work, and the part projects most often rush. At least three families of rules.
Deduplication. Define the uniqueness keys (email for contacts, domain for companies, plus edge cases: subsidiaries, multiple domains, contacts linked to several companies), deal with the existing stock of duplicates, then set up the mechanism that prevents new duplicates from coming in: merge rules, import checks, team training. Deduplication is not a technical task. It is a governance decision: who decides which record survives the merge, and on what criteria.
Property dictionary. For each critical property, document its definition, format, allowed values, source of truth and owner. This is where the conflict between declared and enriched data is settled: for each property, you decide whether Breeze can write, in which cases (empty field only, or overwrite), and who arbitrates discrepancies.
Normalization. Phone formats, name capitalization, value lists (industries, countries, company sizes) aligned between HubSpot and the rest of the information system. If the CRM is synchronized with an ERP, this mapping work determines everything: HubSpot is not an ERP, and a sound architecture leaves each system its role (business complexity to the ERP, customer interactions to the CRM) with explicit mapping rules between the two. Our CRM integration team spends a significant part of its projects on this topic alone.
Once the database is governed, activate Breeze scope by scope rather than all at once. Start with enrichment when new records are created (the flow), before processing the historical stock in prioritized batches: target accounts first, dormant database next, if the value-to-credits ratio justifies it. Measure for each batch: completeness rate before and after, rate of divergence from existing data, credit consumption. Only then open up buyer intent and form shortening, which need the enriched data to be reliable in order to produce value.
This gradual approach is not excessive caution. It protects credits, makes it possible to correct priority rules on a limited scope, and gives teams time to see that the data is improving. Trust is rebuilt through evidence.
Enrichment is not an end in itself. Here is what it unlocks, team by team, once governance is in place.
Marketing ops. Form shortening reduces the fields displayed on conversion forms without reducing the data collected: the completion rate improves, segmentation remains possible. Enriched firmographic data makes lead scoring more reliable (scoring based on fields that are 60% empty scores nothing) and enables segmentation by size, industry or technology stack that was impossible with declared data alone.
Sales and SDRs. Buyer intent turns anonymous web traffic into a list of accounts to work, prioritized by fit with the ideal customer profile. Enriched records save pre-call research time: headcount, news, technologies in place. In a B2B sales cycle where SDRs spend a large share of their time qualifying, reliable data upfront changes the nature of the first conversation.
RevOps. This is the team that gains the most, and the least visible one. Territories and routing based on complete headcount and industry data, forecasts based on clean segments, industry reporting that is finally usable. RevOps is also the natural owner of the governance rules described above: arbiter of source priorities, guardian of the property dictionary. Our SalesOps and data management offer covers exactly this scope.
Customer service. Less central, but real: a complete company record when a ticket arrives makes it possible to prioritize according to the account's size or segment, without switching between tools.
Let us be honest about the limits, because they determine the return on investment.
Coverage first. The Breeze database is richer for some geographic areas and company profiles than others; the match rate on a file of French SMEs will not be the same as on a file of US scale-ups. Before committing the full stock, test the enrichment rate on a representative sample of your market. Ten minutes of testing avoids weeks of disappointment.
Cost next. The credit model requires reasoning in terms of value per enriched record: enriching 200,000 dormant contacts of which 5% are still active is rarely a good calculation. The orders of magnitude are detailed above: check them against current HubSpot pricing, as they change.
Freshness last. Enriched data ages like entered data. Enrichment is a flow to maintain, not a one-off operation: governance rules must include periodic revalidation of critical properties.
And a limit of scope: Breeze Intelligence completes data, it does not clean it. Duplicate detection, format normalization and archiving dead records remain separate tasks, handled with different tools (native HubSpot features, controlled import operations, sometimes dedicated tooling). Believing that enrichment will clean up the database is the most common misconception we encounter in pre-sales.
The first decision is not to activate Breeze. It is to appoint an owner for CRM data, with a clear mandate and dedicated time. Without this role, the audit will lead nowhere, the rules will stay in a document nobody applies, and enrichment will produce volume instead of reliability.
The second decision is to size the governance work honestly before budgeting for the tool. In most of the portals we audit, cleaning and rules represent most of the effort; activating Breeze Intelligence, when properly prepared, is the short part of the project.
Then come the agents. Not before.