How And When To Use HubSpot Smart Properties

A team we work with recently ran HubSpot’s smart properties across their entire company database. Around 3,000 records. The feature looked sharp in the demo. Simply enter a prompt, pick a data source, and let AI fill in the gaps. Thirty minutes and 30,000 credits later though, roughly a third of the results came back blank. Another third were vague enough to be useless, and the remaining third were genuinely good.

The feature worked, but the infrastructure underneath it, didn’t. We see this sort of thing time and time again. While smart properties are one of the more useful Hubspot AI features, most AI tools are only as good as the foundation they sit on. And most of the guides about these tools seem to assume the hard bit is what to click on, skipping the infrastructure part entirely and jumping straight to the same setup steps you’ll find in product documentation.

This is a different kind of guide because it covers what smart properties are, where they earn their keep, what they cost at scale, when to use something else, and what needs to be accurate in your CRM before any of it works.

What HubSpot smart properties are

Smart properties are CRM fields that use AI to populate themselves. You write a prompt, choose a data source, and HubSpot’s Breeze engine fills in the value for each record. They sit inside the Data Agent, under Data Management.

The four data sources you can pull from:

• Web research. Breeze searches publicly available information using the Company Domain Name field. Good for industry classification, company summaries, and tech stack detection.
• Company website. Pulls directly from the company’s own site. Useful for extracting services offered, target markets, or positioning language.
• Property data. Reads from an existing property on the same record. Lets you derive or reformat values you already have.
• Call transcripts. Analyses the last five recorded calls on a record. Useful for surfacing objections, summarising next steps, or flagging risk.

Smart properties are available on contacts, companies, deals, tickets, and custom objects. You need a Starter plan or above on any Hub, plus Edit property settings or Super Admin permissions to create them.

Here is where people get confused. There are three different ways to automatically set a property value in HubSpot, and they do different things:

Tool	What it does	Best for	Cost
Smart properties	AI enrichment from prompts and external data sources	Research, classification, summarisation	10 credits per record
Calculated properties	Formula-based maths on existing fields	Scores, totals, date calculations	Free
Workflow property actions	Set values based on triggers and conditions	Lifecycle updates, routing, tagging	Free

• If you need maths, use a calculated property
• If you need a value set when something happens, use a workflow
• If you need the AI to go and find something, or make sense of unstructured data, use smart properties

This distinction matters because using the wrong tool wastes credits on a job that costs nothing with the right one.

Where smart properties earn their credits

Smart properties are worth the cost when they replace manual work that is either not happening or happening inconsistently. The best use cases for it generally involve high-value records, a specific prompt, and a clean data source.

Pre-call research. Before a sales call, you want to know what the company does, what they sell, and what their tech stack looks like. A smart property using “web research” can pull a summary in seconds. The specificity of the prompt is key; “Summarise what this company does in one sentence, based on their website” returns something useful, whereas “tell me about this company” is far less likely to.

A prompt like this works:

“Based on the company website, describe the company’s primary product or service, their target market, and their approximate size (SME, mid-market, or enterprise). Return as three short bullet points.”

The output is not perfect but it is consistently better than the nothing that was there before, and it saves a rep 10-15 minutes per record.

ICP fit scoring
You can use property data to classify whether a company matches your ideal client profile. Write a prompt that references the properties you care about (industry, headcount, revenue band, tech stack) and ask Breeze to return a fit score or category. This turns a subjective gut check into something repeatable.

Call transcript intelligence
If your team records calls in HubSpot, smart properties can summarise the last five conversations on a record. “What were the main objections raised in recent calls?” is useful for deal reviews, handovers between reps, and churn risk detection.

Lead qualification
For inbound leads, a smart property can classify contacts into segments or personas using form data and web research. The prompt defines the categories; the AI does the triage. This works well when the volume of inbound leads makes manual classification impractical.

What smart properties cost at scale

Each smart property fill costs 10 HubSpot Credits. You pay whether the result is useful or not. If the data source is empty or the domain is wrong, Breeze returns nothing and still charges.

Your monthly credit allocation depends on your plan:

Plan tier	Credits per month
Starter	500
Professional	3,000
Enterprise	5,000

Additional credits come in packs of 1,000 at $10 each. This is the bit that catches people out, because if your usage triggers a tier upgrade, that increase is added to your bill permanently. It is not a one-off charge.

One detail easy to miss is that all AI features share one credit pool. Smart properties, the Prospecting Agent, the Customer Agent, and Breeze workflow actions all draw from the same monthly allocation. Heavy smart property usage in marketing means fewer credits for the sales team’s AI features, so decide who owns the budget before someone burns it on day one.

The maths at scale is straightforward. 5,000 company records enriched with one smart property = 50,000 credits, roughly $500 USD. Run three smart properties across the same set and you are looking at $1,500 USD. One bulk operation, one click.

Do not use the bulk fill button at setup. When you create a smart property, HubSpot offers to fill it across your existing records right there in the setup flow. Decline it. Create the property empty. Then enrol records through a workflow instead, so you control which records get enriched and at what pace. The bulk-fill button is exactly how expensive mistakes happen.

Enrich selectively, not across the whole database. A three-tier approach works well:

• Always enrich: MQLs and named target accounts. These are the records where ten cents of research drives a decision. Spend freely here.
  
• Conditionally enrich: Accounts showing fresh engagement. They have crossed a threshold worth a closer look, but are not automatic. Gate these behind a workflow trigger.
  
• Do not enrich: Cold records, out-of-ICP companies, and anything you would disqualify on a deterministic rule. Researching a record you would never work is setting money on fire at volume.

When smart properties are the wrong tool

Every guide on smart properties tells you how to use them but almost none tells you when not to. Which is the more useful question.

1. When your Company Domain Name field is empty or wrong. Smart properties using “web research” depend entirely on this field. If it is blank, Breeze has nothing to search. If it is wrong (a LinkedIn URL instead of a domain, or an old domain that redirects), the result is garbage. You pay the credits either way. Before you run smart properties at any scale, audit this field. It is the single biggest predictor of whether enrichment will work.
   
2. When a workflow does the job for free. If you are setting a property value based on a trigger or a condition, a workflow costs zero credits and executes instantly. “When lifecycle stage changes to SQL, set Lead Priority to High.” That is a workflow, not a smart property. Using AI for logic is like hiring a consultant to do your filing.
   
3. When a calculated property is what you need. Whenever the answer is a scalculation like sum of deal amounts, days since last activity, weighted lead score, use a calculated property. Smart properties cannot do arithmetic, they are designed for interpretation and summarisation, not computation.
   
4. When the volumes do not justify the cost. For small, targeted enrichment (a segment of 100-500 high-value records), smart properties are efficient. But for ongoing, bulk enrichment across tens of thousands of records, third-party tools like Apollo or Clay are often cheaper and deeper. Do the maths before you hit “Fill smart properties” on a large segment.

We wrote about the demo trap before: buying based on what a feature can do in ideal conditions, rather than what it will do in yours. Smart properties at scale are a textbook case.

How to set up a smart property that works

The enrichment quality depends almost entirely on the prompt and the data source:

Step 1: Navigate to Data Management > Data Agent. Click “Create smart property” in the top right.

Step 2: Choose your object and field type. Select the object (contact, company, deal, ticket, or custom object), name the property, and choose a field type. Single-line text and dropdown are the most common.

Step 3: Write your prompt. This is where most people get it wrong. A vague prompt produces a vague result. Be specific about what you want, what format you want it in, and what the AI should do when data is missing.

One rule worth following is to pin the output to exact allowed values. The agent will happily write you a paragraph when you wanted one word. Define the only acceptable outputs and forbid everything else. If you want a rating, give it a controlled vocabulary. If you want a score, specify the range. This is what makes smart property outputs filterable, reportable, and usable in workflows.

A good prompt:

“Using web research, identify the primary industry this company operates in. Return one of the following categories: Professional Services, Financial Services, Technology, Education, Manufacturing, Healthcare, Other. If the industry cannot be determined, return ‘Unknown’.”

A bad prompt:

“What industry is this company in?”

The difference is specificity. A good prompt defines the output format, gives the AI a controlled vocabulary, and handles the edge case. Whereas a bad prompt leaves everything open and produces inconsistent results you cannot filter or report on.

Step 4: Choose your data source. Match the source to the question. Web research for external facts. Company website for positioning and offerings. Property data for deriving values from existing fields. Call transcripts for conversation intelligence.

Step 5: Test on 2-3 known records. Before you run a smart property across a segment, test it on records where you already know the answer. If the output does not match what you know to be true, refine the prompt before scaling.

Step 6: Run on a segment. Navigate to CRM > Segments, select a segment that matches the object type, click Actions > Fill smart properties. Watch the credit consumption.

Permissions matter here. Only users with Edit property settings or Super Admin access can create smart properties. Restrict this. An enthusiastic team member with bulk enrichment access and no credit awareness can do real financial damage in minutes.

Use workflows, not manual fills. Here’s three sequences worth building:

• New MQL enrichment. Trigger: contact reaches MQL lifecycle stage. Filter: company is in-ICP and has a valid domain. Action: fill your smart properties on the associated company record.
  
• High-intent engagement refresh. Trigger: contact views a pricing or demo page, or 3+ contacts from one company engage within 7 days. Filter: company not enriched in the last 14 days. Action: refill engagement and intent properties.
  
• Target account quarterly refresh. Trigger: scheduled, quarterly. Filter: company is on the named target-account list. Action: refill all web-research properties so the data stays current on the accounts that matter.

These ideas come from Ryan Gunn’s excellent prompt library for smart property lead scoring, which is worth reading if you want copy-paste-ready prompts for ICP scoring, LinkedIn activity indexing, and sales cycle estimation.

Smart properties vs third-party enrichment tools

Smart properties are native to HubSpot whereas third-party enrichment tools (e.g. ZoomInfo, Clay, Apollo) sit outside it and push data in. The right choice depends on what you need.

Factor	Smart properties	Third-party enrichment
Data depth	Surface-level web research. Good for summaries, classification, and interpretation	Deeper firmographic, technographic, and intent data. Contact discovery, org charts, funding data
Cost model	10 credits per record per fill. Baked into HubSpot billing	Subscription or per-record. Separate contract and budget line. Requires integration, mapping, and ongoing maintenance
Integration	Native. No setup required	Requires integration, mapping, and ongoing maintenance
Best for	Selective enrichment of high-value records. AI-powered classification and summarisation	Bulk enrichment. Contact discovery. Deep firmographics at scale
Main risk	Credit burn on dirty data. Shallow results for niche industries	Data staleness. Integration breaking on HubSpot updates. Vendor lock-in

For most mid-market B2B teams, it makes sense to use smart properties for targeted, AI-driven enrichment on records that matter, and and a third-party tool when you need depth, volume, or data types that HubSpot’s web research cannot reach.

Getting the foundations right before you turn anything on

Smart properties are only as good as the CRM they sit on. If your data model is inconsistent, your Company Domain Name fields are half-empty, and your lifecycle stages are undefined, no amount of AI enrichment will produce reliable results.

Before you create your first smart property, check these:

• Company Domain Name. Populated and correct across your database. This is the field web research depends on, so audit it.
  
• Lifecycle stages. Defined and consistently applied. Smart properties that reference lifecycle data are useless if the stages are meaningless.
  
• Property naming conventions. Consistent and documented, so adding AI-generated
  properties to a CRM with 400 unnamed custom fields only makes the mess worse.
  
• Permissions. Restricted to people who understand the credit implications. This is not a feature to democratise on day one (if ever).

Start small with one use case and one segment of high-value records. Run it, check the output, and measure whether it saved time, improved data quality, or surfaced something actionable. If it did, expand. If if it didn’t, figure out how to fix the foundation before trying again.

What we commonly see is that these features are ready (and ready to have you spend with them), but the infrastructure that you control beneath it probably isn’t. That gap between what your CRM can do and what it currently delivers is where the work is. And it is usually an architecture problem involving data models, governance, process, and decisions that might have got deferred when the platform was set up.

If any of that sounds familiar, that is the kind of CRM optimisation Fly does through our Flight Plans. We rebuild the foundations so the features on top of them work, then keep them working and evolving as your business scales.

---

Smart properties are a good feature, but on a well-built CRM they are a great one. Getting the foundations right is the challenge. If that’s something you need help with, talk to Fly.