If you were an early adopter of custom Breeze Assistants in HubSpot, something has shifted under you this month. As of 13 July, existing custom assistants became read only, automatically migrated into Breeze projects, and new custom assistants stopped being creatable a few weeks earlier on 19 June.
On the surface that reads as a routine sunset, but it's more interesting than that. What HubSpot has actually done is quietly reshape Breeze to look a lot like the shared, context aware workspaces we've been seeing ChatGPT and Claude build with their own Projects features. It's a small change on top and a much bigger one underneath about how HubSpot expects teams to work with AI inside the CRM.
What custom Breeze Assistants were, and why they were useful
A custom Breeze Assistant used to be a self contained thing. You configured it once in Breeze Studio, gave it a welcome message and a few conversation starters, attached the tools and knowledge it should draw on, and it lived as its own object inside your portal. That worked well for narrow, repeatable tasks like a support triage helper, a specific content brief generator, or an internal Q&A bot on a particular process. The benefit was focus: one job, one configuration, one place to go for that job.
The limitation was the same as the benefit. Every assistant was a silo. If a workstream evolved or picked up new context, you were either reconfiguring the assistant or spinning up another one. And because assistants sat separately from the rest of the way people actually work, the context tended to live inside your head between chats rather than in HubSpot.
What's replaced them: Breeze projects
Breeze projects work differently. A project is a shared workspace that groups related chats under one roof, with three layers of grounding underneath every conversation inside it.
-
At the top, there are the project's own instructions and reference files, uploaded directly or pulled in via @mentions of CRM records. That's the closest equivalent to what a custom assistant used to hold.
-
Underneath that sit Knowledge Vaults, which are still very much in play. Vaults are curated, controlled sets of context (files, CRM segments, knowledge base articles) that you manually attach for specific use cases (up to 50 per portal, permissioned per vault, and unchanged in purpose by this migration).
-
And underneath that is Context, a newer layer available under Breeze > Context. Context is foundational business information that HubSpot applies automatically across every AI feature in the portal: your brand kit and voice, your business profile, target personas, market position, tech stack, products and services, email personality, and so on. You set it once and every Breeze surface uses it.
Layered together, that's automatic business context underneath curated workstream context underneath project level instructions and files. Every chat inside a project inherits the whole stack. You stop re-explaining who you are, what you sell, and what you're working on every time you open a new thread.
Access is different too. Projects are shared team spaces by default with permission levels, so the right people see the right projects rather than each person maintaining their own private assistant.
What this means for your team
If you have custom assistants running, there's a short list of things worth doing this week.
Audit which custom assistants exist in your portal and who was relying on them. They're now read only, so anything operational needs a plan. Manually rebuild the welcome messages and conversation starters as project instructions, because that step doesn't carry across automatically. Review the permission levels on the new shared projects, since these are team visible workspaces rather than single owner tools.
Then take the slightly bigger step. Look at your Context configuration in Breeze and make sure your business profile, personas, brand kit, and products are actually complete and current. That's the layer doing the quietest work across every AI interaction in your portal, and most teams have never opened it. Then think about your projects the way you'd think about your workstreams: a project per campaign, per key account, per RevOps initiative, per sales play, each with its own instructions and its own set of Knowledge Vaults attached. That's the shape Breeze is now built for.
Where this sits in the wider AI landscape
HubSpot isn't inventing anything radical here. ChatGPT rolled out Projects to group chats with shared instructions and files. Claude did the same. What's genuinely worth noticing is that three companies, working from very different starting points (a horizontal chat product, a horizontal chat product, and a CRM), have all converged on the same shape: a persistent workspace that holds your context, your files, your instructions, and increasingly your team, so that AI has something to work from instead of starting cold every time. The unit of AI work is becoming the project, not the prompt.
Where HubSpot is arguably ahead is on the layering. ChatGPT and Claude Projects mostly stop at project level instructions and uploaded files. HubSpot has that layer, plus curated Knowledge Vaults, plus automatic business Context underneath everything, plus live CRM records available via @mention. If you're grounded in the CRM already, that's a materially richer stack.
The bigger observation: HubSpot as workspace, not just CRM
The direction of travel is worth naming. HubSpot has spent the last few years pulling in work that used to happen outside it. First it was more of the marketing stack, then commerce and billing under the Revenue Hub rename, and now the AI work most teams have been doing in ChatGPT or Claude or standalone assistants. Bit by bit, HubSpot is turning itself into the place your team actually works, not just the place your customer data lives.
That reframes a lot of decisions. If your Breeze projects and Context are set up well, the AI drafting, prospecting research, meeting prep, and campaign planning your team is currently doing in half a dozen tools can move into HubSpot, sitting on top of the CRM data those tasks are supposed to be grounded in anyway. That's a real shift in how a team's day looks. Whether it's the right shift for every business is a separate question, but the direction is clear.
Ready to review your processes?
If you had custom Breeze Assistants running and want a hand auditing what needs rebuilding as projects, or you're looking at Context and Knowledge Vaults as a first proper AI foundation for your team, that's exactly the kind of setup work we do every week as a HubSpot Elite partner. Get in touch and we'll walk through it with you.
Frequently asked questions
Do I need to do anything if I had a custom Breeze Assistant?
Depends - existing assistants are now read only, and their welcome messages and conversation starters need to be manually rebuilt as Breeze project instructions. The assistants themselves have all migrated automatically to projects - check your set up and see if you need to make any changes.
Are Knowledge Vaults being retired too?
No. Knowledge Vaults are still active and still the recommended way to attach curated context (files, CRM segments) to specific projects, assistants, or agents.
What's the difference between Context and Knowledge Vaults?
Context is foundational business information applied automatically across every HubSpot AI feature. Knowledge Vaults are curated context sets manually attached for specific use cases. Most teams should use both.