Skip to content
Book a build

How do I use AI to audit a HubSpot portal?

To use AI to audit a HubSpot portal, start from HubSpot's data quality tools, then give the AI structured data, not screenshots: property usage, duplicates and a workflow inventory. Claude and ChatGPT connectors can read CRM data but not workflow definitions, so they miss the automation layer. Ask for findings with evidence, and score against fixed rules rather than the model's opinion.

On this page

What HubSpot supports natively

HubSpot’s own audit-like tooling is the Data Quality page (Data Management → Data Quality). As of its July 2026 article it shows:

CheckWhat you get
DuplicatesContact and company duplicates to merge or reject
Formatting issuesValues that don’t match the expected format
Property insightsDuplicate, empty and unused properties (anomalies need Data Hub Professional or Enterprise)
Unused workflowsA list of workflows you aren’t using
EnrichmentCoverage and gaps for contacts and companies

The page is available on Starter, Professional and Enterprise editions of Marketing, Sales, Service, Data and Content Hub, and on Professional and Enterprise Smart CRM and Revenue Hub.

What it doesn’t do: explain how workflows interact, find workflows that set the same property in opposite directions, or score the portal against a standard. That’s where people reach for AI.

The AI connectors don’t close that gap on their own. HubSpot’s connectors for Claude and ChatGPT read CRM records, marketing assets, properties, pipelines and saved reports, but neither lists workflows, so a chat-based audit sees the data and not the automation that changes it.

How to use AI to audit a HubSpot portal

  1. Pick the scope. Data model, automation, sales process, marketing assets or reporting. One area per pass gives better findings than “audit everything”.
  2. Collect structured inputs. Property definitions and fill rates, duplicate counts, pipeline stages, and a workflow inventory with triggers, actions and the properties each one reads or sets.
  3. Fix the rules before the model reads anything. For example: a property unused for 12 months, a workflow with no enrollments in 90 days, two workflows writing the same field.
  4. Ask for findings with evidence. Each finding names the asset, the data behind it and a recommended action. Reject findings that cite nothing.
  5. Verify the top ten by hand in HubSpot before you report them.
  6. Keep fixes separate. An audit reads; changes go through a reviewed plan.

Common mistakes

  • Letting the model invent a health score with no rule behind it.
  • Auditing workflows from names and descriptions instead of their definitions.
  • Giving the auditing tool write access “in case it can fix things”.

If you’d rather hire it out, see how to choose an AI HubSpot audit service.

Where Jetstack AI fits

Jetstack AI’s Portal Audit does the whole method above for you, in minutes. It reads the entire portal and scores it with fixed rules: 35 blocks across six sections (general, sales, marketing, service, automation and reporting) and more than 700 data points, each with a value, a status and a recommendation. AI writes the narrative on top of the scored data, never instead of it. The Workflow Audit goes deeper on automation: it maps the full workflow inventory, dependencies and enrollment paths, flags property collisions, enrollment overlaps, circular chains, orphaned dependencies, dead branches and re-enrollment risks, and scores every workflow. Both only read your portal, so you decide what to change with the full picture.

To audit from Claude or ChatGPT, connect Jetstack’s MCP server. Your assistant can run a Portal Audit on any connected portal, read back every block’s data points, recommendations and health scores, and pull any workflow live from HubSpot as a diagram to reason about. Workflow Audit results go to your own scripts and agents through the Workflow Audit API. To build the fixes an audit recommends, Projects (currently in limited release) builds a corrected version as a new asset from a plan you approve, with workflows created switched off, so you can retire the old one in HubSpot.

Sources

Checked October 30, 2026. HubSpot changes its products often; tell us at team@jetstack.ai if something here is out of date.

AI, Claude & MCP

See what needs work before you change it.

A read-only Portal Audit and Workflow Audit score the portal, trace dependencies and turn findings into a branded report.

Explore Portal Audit

AI, Claude & MCP

Related questions.

How do I use AI to audit a HubSpot portal?

To use AI to audit a HubSpot portal, start from HubSpot's data quality tools, then give the AI structured data, not screenshots: property usage, duplicates and a workflow inventory. Claude and ChatGPT connectors can read CRM data but not workflow definitions, so they miss the automation layer. Ask for findings with evidence, and score against fixed rules rather than the model's opinion.

Can Claude audit my HubSpot workflows through the connector?

Not fully. HubSpot's Claude connector doesn't list workflows among the objects it can read, so Claude can reason about properties and records but not about what your workflows do. A workflow audit needs the workflow definitions, which come from the Automation API.

Is an AI audit accurate?

As accurate as its inputs and rules. Counts and checks (unused properties, workflows with no enrollments, missing owners) should come from data; the AI is useful for explaining findings and prioritising them. Treat an AI score with no stated rule behind it as an opinion.

Does an AI audit change anything in HubSpot?

It shouldn't. Run audits with read-only access where you can, and keep any fixes as a separate, reviewed step.