Structured CRM data lives in defined fields, like lifecycle stage, deal amount, or industry, so software can filter, count, and report on it. Unstructured CRM data is free-form content, like emails, call transcripts, notes, chats, and files. AI needs both. Structured fields tell it who a record is and where it stands, and unstructured content tells it why, in the customer's own words.
Breeze can only work with what your portal holds. It gives weak answers when key fields are blank, and it misses the real story when calls and emails never reach the record. If you're planning to roll out HubSpot agents, our HubSpot AI agent implementation team prepares the CRM data and knowledge sources each agent depends on.
Structured CRM data follows a fixed format, and unstructured CRM data doesn't. A dropdown property holds one value from a set list, so HubSpot can segment, report, and trigger workflows on it. A call transcript holds whatever the customer said, so a person or an AI model has to read it before anyone can act on it.
Some data is technically structured but has no standardized values. This includes single-line text properties, legacy tags, and email subject lines. Treat them as unstructured data because variations like "Healthcare," "Health care," and "Healthcare services" appear as separate values in reports.
Unstructured data in HubSpot is any content AI has to read before it can use it. Most portals hold seven kinds.
How much of it AI can use depends on two things. First, transcription analysis needs a Sales Hub or Service Hub Professional or Enterprise subscription, and only calls by users with those seats get transcribed automatically.
Second, a transcript only helps when it sits on the right contact, company, deal, or ticket. On those tiers, HubSpot AI can link calls to deals and tickets automatically, so agents can find the conversation.
AI needs both, and each type does a different job. Structured data tells AI what records to look at and how they compare. Unstructured data explains the reasons behind the numbers.
In 2025, HubSpot relaunched Operations Hub as Data Hub and updated Smart CRM to bring structured, unstructured, and external data together. Gartner predicts that 60% of AI projects will be abandoned by 2026 due to a lack of AI-ready data.
Start by fixing your structured data. AI agents use CRM properties to identify which records to work on. Missing or incorrect property values can send even well-informed agents to the wrong accounts.
HubSpot AI reads unstructured data through four controls. These are your AI settings, Conversation Intelligence, Data Agent, and knowledge vaults.
Under Settings > Account management > AI, CRM data and Customer conversation data are on by default, and Files data starts off.
Customer conversation data covers call transcripts, emails, and similar conversations. With Files data off, Breeze can't use the playbooks and product sheets your team stored in the files tool.
Conversation Intelligence transcribes calls so AI can summarize them and extract details. Turn on call recording and transcription under Settings > Calling > Call Setup on the Call Configurations tab, and connect Zoom or Google Meet if your team runs meetings there.
Data Agent fills smart properties. These are AI-generated fields that pull from web research, the company website, property data, or call transcripts. In workflows, the Data Agent research action can use the five most recent call transcripts on an enrolled record. It returns text, a number, or a true or false value.
Each run uses HubSpot Credits for every prompt on every record, so start with the properties your team will actually filter or report on. Smart properties can't use validation rules, so review a sample of filled values before you build workflows on them.
Knowledge vaults give specific agents extra context, such as PDFs and other documents, HubSpot content, or CRM segments. You can keep up to 50 knowledge vaults at once and attach each one to the agents that need it. Together with your CRM records, these sources build your HubSpot Growth Context, the business knowledge HubSpot's agents work from.
Turn unstructured data into a property when you need to filter, report, automate, or score on it. Leave it as text when people or AI only need to read it for background.
We used the same rule in the Fullstaq Keap to HubSpot migration. Fullstaq had more than 4,500 legacy Keap tags. We rebuilt them as HubSpot properties and lists, so three business units could segment and report from one shared set of fields.
Messy values cause trouble even inside structured fields. In our NETGEAR reseller data foundation project, legacy picklist values blocked routine Salesforce sync updates until we reconciled them. After the data foundation work, the reseller list's email block rate dropped from 80% to 4.6%.
Prepare structured data by cleaning the properties AI filters on, and prepare unstructured data by making sure conversations get captured on the right records. Follow these steps.
If your customer data lives in several systems, our HubSpot Data Hub guide walks through the sync and cleanup tools.
An agent can only act on the properties and conversations it can read. Our AI agent process starts with a Prepare stage that organizes the knowledge sources and CRM data each agent needs. Then we configure and test each agent and train your team.
As a HubSpot Elite Solutions Partner with accreditations in CRM Implementation, Data Migration, and Solutions Architecture, we build the data foundation first. If you're getting ready to launch Breeze agents, talk to our HubSpot AI agent implementation team.
Structured CRM data tells AI who and what to act on, and unstructured data tells it why. HubSpot AI can read both, but only when properties are clean, conversations are captured, and the right AI settings are on.
Check how ready your portal is with the free HubSpot AI Readiness Index, a 12-question diagnostic that takes about 3 minutes.