Updated October 8, 2026.

How does HubSpot’s Data Agent work?

HubSpot’s Data Agent uses prompts and configured sources to research, transform, and populate information. Its capabilities include Smart Properties, Smart Actions, and Smart Columns. Smart Properties can use supported sources such as web research, record properties, and activities or transcripts. The available inputs and behavior depend on the configuration.

Sources: Set up and use Data Agent and Create Smart Properties.

For businesses using HubSpot, the practical opportunity is to turn a useful research question into information the team can work with.

Instead of repeatedly asking someone to investigate an account, define the question, choose the appropriate source, test the answer, and decide how the result should support your process.

The quality of that definition matters as much as the decision to use AI.

Smart Properties, Smart Actions, and Smart Columns

HubSpot groups Data Agent capabilities into three areas:

Source: HubSpot Data Agent overview.

Choose the capability that fits where the information needs to be used. A result intended for ongoing account qualification deserves a different design from an exploratory analysis.

Before configuring anything, answer:

• What business question are we asking?
• Which source can support the answer?
• Where should the result appear?
• Who will use it?
• What should happen when the evidence is incomplete?

How do Smart Properties work?

Smart Properties pair a property with a prompt and a configured data source. HubSpot supports previewing the result on a selected record before filling properties.

Source: Automatically populate records using Smart Properties.

Start with a question that supports a real decision.

For example, a team selling into manufacturers might want to understand whether a target company operates multiple locations. A fundraising team might investigate a defined public signal relevant to donor research.

These are potential use cases to evaluate. The right configuration depends on the source, available evidence, property format, and intended use.

A populated field should make someone’s next step clearer.

Write prompts with a clear definition of success

A broad instruction such as “Is this company a good fit?” leaves important business rules unstated.

A more useful hypothetical prompt could be:

“Using the selected source, determine whether the company publicly describes manufacturing operations. Return Yes, No, or Unknown. Use Unknown when the evidence is insufficient.”

Adapt the wording to the property type and the options supported by your configuration.

Test whether the prompt distinguishes a negative finding from missing information. “No evidence found” should not automatically become “No.”

For more complex research, separate questions rather than forcing several judgments into one property.

Choose the source deliberately

HubSpot’s Smart Property documentation describes source options including web research, company websites, property values, and activities or transcripts.

Source: Smart Property configuration.

Match the source to the question. A public company announcement and an internal customer conversation provide different evidence.

Do not assume that connecting an ERP, billing platform, or other system makes every field available to every Data Agent prompt. Verify the specific information and supported source configuration.

If operational information is required, confirm the integration, record associations, and usable properties before testing.

A practical Smart Property setup process

Use this sequence to plan your configuration:

1. Define the question. Explain what decision the result will support.
2. Choose the property format. Decide how the result should be represented.
3. Select a supported source. Confirm it contains the information needed.
4. Write the prompt. Include criteria and how to handle uncertainty.
5. Preview representative records. Test clear cases and incomplete information.
6. Review evidence and outputs. Check whether the results meet your criteria.
7. Fill a limited set of records. Evaluate quality before increasing volume.

HubSpot documents creating Smart Properties through record views and property configuration. It also provides source details for filled values.

Source: Create and review Smart Properties.

Choose test records that expose weaknesses. Include incomplete domains, similarly named companies, sparse activity history, and conflicting information where relevant.

A prompt that works on one ideal example is not yet ready for a broad rollout.

How do Smart Properties fit into workflows?

HubSpot documents the Data Agent: Fill Smart Property workflow action.

Source: Fill Smart Properties through workflows.

Design the process around the value of the result.

For a hypothetical account-research workflow:

• Select a defined group of records.
• Populate the research property.
• Inspect uncertain or unexpected results.
• Route useful findings to the appropriate owner.

Confirm the supported workflow actions, subscription requirements, and review behavior before implementation.

Use narrow enrollment criteria during the pilot. Decide how existing property values should be treated and whether repeated processing is appropriate.

How should you review Data Agent results?

Review both the populated value and the evidence supporting it. HubSpot provides source details and Smart Property management metrics, including successful runs and errors.

Source: Review sources and manage Smart Properties.

A successful run is a technical outcome. It does not establish that the result is useful or correct.

Use a review checklist:

Record recurring errors so you can distinguish source problems from prompt problems.

What does Data Agent cost?

HubSpot’s current catalog lists 10 credits for one Data Agent prompt response for one record. At $0.01 per additional credit, that represents $0.10 in purchased-credit usage value.

Source: HubSpot Credits Rate Sheet.

For a hypothetical batch of 200 records, with one response each:

200 × 10 credits = 2,000 credits, or $20 of purchased-credit usage value.

That calculation is not the final invoice. Included credits, billing settings, repeated requests, and other workflow consumption affect the overall cost.

Check the rates for every component of a larger process rather than treating the Data Agent response as the entire workflow cost.

Read HubSpot Breeze Pricing for budgeting guidance.

Data readiness still matters

Before expanding the use case, review:

• Record identity and duplication.
• Relevant property definitions.
• Required source information.
• Record associations.
• Ownership and downstream actions.

Focus on the information needed for the chosen task. Your entire CRM does not need to be perfect before a controlled pilot, but the inputs that drive the process need deliberate review.

For more guidance, read Do I Need Clean CRM Data Before Using Agent Hub?.

Measure useful research, not just populated fields

Track the difference between activity and value.

Useful measures include:

• Accuracy across a reviewed sample.
• Results requiring correction.
• Missing or uncertain answers.
• Research time recovered.
• Credit consumption.
• Whether the information improves a defined decision.

If a property fills consistently but nobody uses it, revisit the purpose before increasing volume.

If the information supports better qualification, preparation, or routing, document that benefit and assess the cost of maintaining it.

Fast Slow Motion’s approach

At Fast Slow Motion, we start by defining the business question and the process it supports.

Then we identify usable information, configure the research, test the results, and establish visibility into quality and cost.

That gives Data Agent a practical role: producing information your team can evaluate and use within a clearly owned process.

Need help choosing Data Agent use cases or configuring a pilot? Talk with Fast Slow Motion about turning research questions into useful CRM information.