HubSpot

Is your CRM ready for agents? A practical look at Context Home

Manjit Gill 6 min read

On 16 September HubSpot launched Context Home: one place to see and manage what its AI knows about your business, with a score that shows how complete that context is and where the gaps are. It matters because Breeze Assistant, Prospecting Agent and custom agents all draw on it. But the score is not a CRM audit, and it does not decide when an agent is ready to go live: that is still up to you.

What Context Home is

HubSpot's Fall 2026 Spotlight introduced Context Home as the place that "gives you a score that shows how complete your context foundation is and where the gaps are". The product page puts it more plainly: "Context home is where you manage what HubSpot knows about your business".

It is new, generally available and included in all HubSpot editions, with no separate charge listed. HubSpot keeps it current in two ways: it "automatically syncs context from your connected sources as your business data changes", and it will "surface recommendations when your context may be outdated". Anything you edit yourself takes precedence over what it syncs.

Breeze Assistant, Prospecting Agent and custom agents use it, along with other agents in Agent Hub. That is the point of it: describe your business, products and customers once, and every agent starts from the same version.

What it holds, and who can change it

HubSpot's knowledge base separates two kinds of information. Context is "foundational information about your business that's automatically applied across multiple AI tools". Knowledge vaults are "sets of information that you manually add to projects or agents for specific use cases".

The Context tab has six sections:

  • Business: identity and classification, business profile, technology stack, market and ecosystem context, brand kit, and products and services.
  • Customers: ideal customer profiles and personas.
  • Team and Processes: your sales methodology.
  • Personal: each user's profile and email personalities.
  • Custom: free-form text and uploaded files, each with its own visibility setting.
  • Knowledge vaults: the use-case sets described above.

You find it under Agents, Agent Hub, then the Context tab, and Breeze Assistant links to it from the panel that shows the context behind each answer. Most of it needs Super Admin or Partner Admin permission to change. Users with Prospecting Agent Admin permission can also edit Products and Services, and everyone manages their own Personal section. In practice that makes Context Home an admin's job, so decide early who owns it and who supplies the words.

What the score tells you, and what it does not

The score is useful as a prompt: it shows where your context is thin and which sections to fill in. But HubSpot has not published how it is calculated, what it weighs or what a good score looks like, and the knowledge base article on managing context does not mention a score at all.

The closest HubSpot comes to a definition is a footnote on the product page: "High-quality context means actively populated CRM fields, logged activity, connected integrations, and configured team settings." Some coverage has described Context Home as a score of CRM data completeness. HubSpot's own documentation does not go that far, and nothing we have found says the score is checked before agents go live, or that agents are held back until you reach a threshold.

So read it as a to-do list for your business context. It will not tell you whether your contact records are complete enough for an agent to answer from them.

Why your CRM data still matters

For the CRM itself HubSpot has a separate tool, the data model health check. It is generally available on all products and plans, under Data Management, Data Model, then the Health tab, and it scans for 16 types of issue in five categories:

  • Objects: custom objects with few records or no associations.
  • Properties: low fill rates, critical properties with low fill rates on record cards, inactive properties, duplicate or similar properties, and properties that are well filled but unused.
  • Associations between records.
  • Pipelines: too many stages, no closed stage, or few records.
  • Lifecycle stages.

A scan takes up to 20 minutes. Issues cannot be dismissed, and HubSpot prompts you to run it again after 30 days. The Fall release also added duplicate similarity scores and merge preview (both generally available), plus primary record merge rules and help with custom duplicate rules (both public beta).

This matters most for an agent that answers from the record. HubSpot's documentation says Customer Agent "can only currently access or update contact properties", up to 10 of them, each with a protection level such as matching the visitor's email address or phone number. If the answer someone needs is not in a well-filled contact property, the agent cannot give it.

The checks before a Customer Agent goes live

Customer Agent is the agent most teams put in front of the public. It needs a Professional or Enterprise subscription and HubSpot Credits, and HubSpot gives it its own pre-launch tools, all generally available in the Fall release:

  • Batch testing runs the agent against real customer questions before go-live, and you can also chat with it in a test window. HubSpot says previewing and testing the agent does not use HubSpot Credits.
  • Deployment recommendations based on your conversation volume and SLA data.
  • Conversation coverage, a percentage that lets the agent start on a share of conversations rather than all of them.

It also will not run until four AI settings are on: "Give users access to generative AI tools and features", "CRM data", "Customer conversion data" and "Files data". Agree those with whoever looks after data protection well before the day you want to launch.

How we close the gaps before an agent goes live

None of this is a single switch, so we work through it in order:

  1. Start with the CRM. We run the data model health check and fix the properties the agent will read first, rather than every issue at once.
  2. Choose what the agent may read. We agree the contact properties it needs, no more than 10, and the protection level for each.
  3. Fill in Context Home. Business profile, products and services, ideal customer profiles, personas, sales methodology and brand kit, written by the people who know them and entered by an admin. Process documents go in Custom, with the right visibility.
  4. Give it content to answer from. Your website and knowledge base first, then short answers for the questions your team gets every day.
  5. Test with your own questions. We batch test with questions taken from your inbox, agree the hand-off rules with your team, and start on a low conversation coverage.
  6. Review after 30 days. We re-run the health check, act on Context Home's recommendations and raise coverage once the answers hold up.

There is one more reason to plan first: once you create a Customer Agent it cannot be deleted, only paused or edited.

If you are planning an agent and want the groundwork done before it goes live, see how we approach AI for HubSpot.

Line up your CRM, conversations and AI.

Tell us what is slowing your organisation down. We will show you what HubSpot, WhatsApp and AI can do about it.

Email
hello@discoverdigital.com
Phone
+44 (0)118 974 4736
WhatsApp
+44 118 974 4736
Office
Reading, UK

Book a call