The HubSpot Data Agent Playbook

The researcher that never sleeps

HubSpot's Data Agent is the part of Breeze that does the research your revenue team would otherwise do by hand. It reads across your CRM, your call recordings, your emails, your documents, and the open web, answers a specific business question you define, and writes the answer straight back into a HubSpot property. HubSpot calls it a customer intelligence agent, and the framing is right: it is not a search bar or an enrichment feed, it is an agent that researches every customer, contact, and prospect at scale and hands your team the answers they used to dig for.

The distinction that matters is what it answers and where the answer lands. Standard enrichment fills a job title or an industry. The Data Agent answers questions like does this company need a sales team, or are they in a buying cycle, the judgment calls that used to require a person opening ten tabs. And because it runs inside HubSpot, the answer does not sit in a chat window; it becomes a property on the record, available to every workflow, list, and report you already run.

This playbook is the complete picture: what the agent actually does, how it turns a question into a CRM field, every source it reads, how it powers workflows and personalization, the setup and the controls that keep it safe, an honest read on pricing and limits, and how INSIDEA deploys it for revenue teams. Every product detail here is checked against HubSpot's own documentation, because the feature is moving quickly.

The problem it removes: the research tax on revenue

Every revenue team runs on research, and almost none of it is counted. A rep preps for a call by reading the account, the last three emails, and the company's recent news. An ops analyst tries to build a segment and realizes the property they need does not exist, so someone has to go find it, one record at a time. A marketer wants to personalize a campaign and hits the same wall: the CRM is full of data, but not the data that answers the actual question.

That is the research tax, and it scales badly. It is fine to research ten accounts by hand; it is impossible to research ten thousand. So teams either spend hours they do not have, or they give up and work off shallow, generic data. The Data Agent is built to remove exactly this tax. It does the reading and the reasoning across every record at once, and it turns the answer into a field you can segment, route, and report on.

The payoff is that the intelligence stops living in people's heads and browser tabs and starts living in the CRM, where the whole team can act on it. HubSpot reports that across an analysis of a thousand real customer prompts, the Data Agent replaced more than 26,000 hours of manual work and automated over 1.1 million research and insight tasks. Your results depend on the questions you ask and the data it can reach, which is what the rest of this playbook is about.

How it works, from question to CRM field

The agent runs a loop, not a one-off lookup. It is worth understanding the whole loop before you configure any of it, because each step maps to a decision you will make when you set it up.

You start with a business question, in plain language, and point it at the records you care about. The agent reads across your sources, your CRM data, your conversations, your documents, and the web, to research an answer. It shows you the results, and, crucially, nothing runs until you approve it. Once you do, it writes the answer back into a property on each record, at scale. From there the answer is just HubSpot data: it flows into workflows, lists, and data studio, and it stays current because you can run the research on an ongoing basis rather than once.

Custom questions, not just enrichment

The single most important thing to understand about the Data Agent is that it is not enrichment with a new coat of paint. Enrichment fills the fields everyone has: job title, industry, company size, the standard firmographics from a fixed data set. Useful, but shallow, and identical for every customer. The Data Agent answers the questions only your team would think to ask, using all of your data.

That difference is the whole point. A firmographic tells you what a company is. A good custom question tells you what to do about it. Here are real examples HubSpot gives of what you can ask, and every one of them is a judgment that used to need a human with a browser and twenty minutes.

Every source it reads

The reason the agent can answer questions enrichment cannot is that it reads far more than a firmographic database. It pulls customer intelligence from across your world and brings it into one place, then reasons over all of it to answer your question.

It reads your CRM records, the structured data you already hold. It reads your conversations, call recordings and emails, which is where most of the real signal lives and where none of it is usually queryable. It reads documents. And it reads the open web, for the things that are simply not in your systems, recent news, a company's own site, public case studies. Bringing those four sources together is what lets it turn a call recording into a set of structured traits, or a company's website into a clear answer about what they sell.

This is also why it belongs to a team on HubSpot rather than a standalone AI tool. The intelligence it produces is grounded in your own data and written back into your own system, so it compounds. Every question you answer becomes a property the next workflow can use, rather than an answer that evaporates when you close the tab.

Research at scale, not one prompt at a time

It is tempting to compare the Data Agent to a general AI chat tool, but the comparison misses the point that makes it valuable. With a standalone tool, you type one prompt about one company, read the answer, and copy it somewhere by hand. That is fine for a single account and hopeless for a database.

The Data Agent researches all your customers at scale, using your existing CRM data, and delivers the answers directly into the CRM, the workflows, and the data studio your team already works in. There are no specialists to hire, no code to write, and no integrations to stand up; you write a question in plain language and it runs across every record you scope it to. That is the difference between an interesting answer and an operational one.

This is the shift from AI as a novelty to AI as infrastructure. The value is not that it can answer a clever question once; it is that it can answer the same question about ten thousand accounts and leave the answers where your team already works.

You stay in control

An agent that writes to your CRM at scale is only useful if you trust what it writes, and HubSpot built the control you need directly into the workflow. The agent does not silently change your data. It previews results before anything executes, and nothing runs without your approval.

That preview step is the habit to build your process around. You see what the agent found before it commits anything to the record, which means you catch a misread question or a shaky web-sourced answer before it lands on ten thousand contacts, not after. Treat the preview as a real review, not a formality, especially on questions that lean on web research or feed a high-stakes workflow. The agent gives you the speed of automated research with a human checkpoint in front of your data, and that combination is exactly what makes it safe to run at scale.

Automate the research, but keep your hand on the data. Preview what the agent found, approve what is right, and only then let it write to the record.

Setup and what you need

Getting the agent working is deliberately light. It is available to HubSpot customers on Starter edition and above, it runs on HubSpot Credits, and it is available in all HubSpot-supported languages. There is no coding, no integration work, and no data team required; you write your question as a simple prompt and the agent does the rest.

The real setup work is not technical, it is editorial. The quality of the property you get out is decided by the quality of the question you put in. A vague question produces a vague field; a precise question, scoped to the right records, produces a field you can build a workflow on. That is where the thinking goes, and it is worth the time.

Powering workflows and personalization

A researched answer sitting on a record is useful. A researched answer wired into your automation is where the agent changes how the team operates. Because the output is a normal HubSpot property, it can drive everything a property can drive.

You can feed the agent's answers into workflows, so records route, score, and trigger based on real intelligence rather than shallow fields. HubSpot customers describe setting up workflows that automatically populate smart properties the moment a call is created, turning every recorded conversation into structured data without anyone lifting a finger. You can build segments on questions that never existed as fields before. And you can power truly contextual personalization: instead of merging in a first name, you can tailor a message to what the agent learned about the account.

This is also how the agent plugs into HubSpot's Loop Marketing model. It supports the Express stage by helping define your ideal customer profile through data analysis, the Evolve stage by surfacing insights for ongoing campaign optimization, and the Tailor stage by enriching profiles with the behavioral context that makes personalization real rather than cosmetic. The through-line is the same: research becomes a property, and the property becomes action.

The use-case library

The agent is not a single-purpose tool, and the teams who get the most from it point it at the questions that actually move their number. Here is the practical library, drawn from how HubSpot's customers describe using it.

Pricing, without the fog

The Data Agent's pricing is simple to reason about: it is available to HubSpot customers on Starter edition and above, and it runs on HubSpot Credits. Some subscriptions include a monthly credit allowance, and you can buy more as you need them. There is no separate per-seat license; the cost is the credits the research consumes.

The practical implication is that cost tracks how much research you run. Answering one question across a small list is cheap; running many questions across your entire database is where credits add up, in the same way that value adds up. That makes it worth being deliberate about which questions are worth asking about which records, rather than researching everything because you can. Always confirm the current credit rates and any included allowance on HubSpot's pricing page, since the credit model can change.

Where it falls short

A comprehensive picture has to include the edges, because knowing them is how you deploy well. The agent is strong at researching and structuring answers and weaker wherever the data is thin or the question is loose.

None of these are reasons to avoid the agent. They are reasons to frame your questions carefully, treat the preview as a genuine review, start with the questions that clearly drive action, and watch your credit usage. Get those right and the limitations rarely bite. Skip them and you will fill your CRM with confident answers you cannot rely on.

How INSIDEA runs it

The agent is powerful, but it rewards clear thinking and punishes vague prompts. The teams who see the reported hours saved are not the ones who asked the most questions; they are the ones who asked the right ones and wired the answers into how they work. This is the sequence we use with clients.

Start from the decision, not the question. Work out what the team would do differently if it knew the answer, because a property that does not change an action is not worth the credits. Frame each question precisely and scope it to the records where it matters. Use the preview as a real review before anything writes to the CRM, and verify web-sourced answers before you trust them at scale. Then do the part most teams skip: wire the new property into a workflow, a segment, or a report, so the intelligence actually drives routing, prioritization, or personalization rather than sitting in a field. Govern the credit usage from day one, and revisit the questions as the business changes.

The agent turns research into a property. Your job is to make sure every property earns its place by changing what the team does next.

This is exactly the work INSIDEA does every day. As an Elite HubSpot Partner, we have built CRM, RevOps, and data foundations for more than 1,500 businesses across 25+ countries, and the pattern that holds up is always the same: a clean CRM, sharp questions tied to real decisions, answers wired into workflows, honest measurement. If you want the Data Agent set up to actually change how your team operates rather than just fill fields, start with a strategy call and we will map the questions worth asking and where the answers should drive action.