When using HS’s Breeze and any AI capabilities, we want to ensure some limits of access.
Is there a way to set certain records as off-limits (for example Companies)? We mainly want to use HS’s AI capabilities for purposes of lead qualification and prospecting, and we want to ensure that there is a limit to existing customer or key account data. Is there a way to say: This list of records should never be included in any AI processing?
What we did do so far, based on what we could find in your documentation:
-Opted out of AI model training in the settings
-Sent an email to request no data being used for machine learning training.
As far as I researched, the “Sensitive data” setting is a way to only set certain Properties as off limits. Does it come down to limiting access of who in the team can use AI tools, as well as setting agent-specific settings?
I’d love to understand this better overall. At the moment, we mainly want to test/use: Prospecting agent, Data agent, Company research agent (potentially), Breeze Assistant + Custom Breeze Assistant.
The relevant settings (which you’ve likely found already) are listed here: Manage AI settings
In general, no, there is no option to set records off-limits - at least not that I know of. HubSpot might at some point introduce this in the future but as far as I know, there aren’t any corresponding settings yet. You could request this in the ideas section of the community: HubSpot Ideas - HubSpot Community
For sensitive data, there’s an important nuance: “Accounts with Sensitive Data turned on are opted out of AI model training by default and can’t opt in.” This just means that AI model training can’t be enabled in portals with sensitive data turned on.
Additionally, here’s how Breeze uses sensitive data: Understand how Sensitive Data is used in HubSpot tools - “some restrictions are in place, which prevent custom property values for Sensitive Data properties from being used to prevent Sensitive Data exposure.”
Besides that, correct, there are instructions you can provide to certain tools (like the agents) and steer them in a desired direction, but this does not limit their data retrieval / usage in any way or form. You can’t tell these features not to look at certain data.
In other words: Overall options aren’t as nuanced currently as you’d like them to be.
To follow up on limiting agent-specific user acces, I am a bit confused and running into an issue. On an example of the Data Agent: I have enabled the Data agent and disabled access of users to this feature.
However, when testing with one of the colleagues (whose access I’ve disabled in the settings) they still have full access to it and can use any of the prompts and even create new ones.
Is there something I am missing? Is the way to limit access only via Settings → User & Teams → edit their access rights as shown in the image?
@Go-T the Data Agent might be something different than what you think. For example, as the permission says, this is referring to smart properties or a specific workflow action. It’s not the same thing as chatting with Breeze in the right sidebar. Could you confirm which AI feature exactly you’re referring to? (What’s the “it” in “they still have full access to it”?)
I understand it’s not the same as Breeze Assistant.
I am referring to Data Agent, which can be accessed via the Data Management section on the left sidebar. As it uses credits and can be used to create things like smart properties or workflows that automatically fill out those properties.
Until testing is finalized, I’d prefer to limit access of users to the Data Agent, which, as shown in my printscreen above should be possible?
If I am getting something wrong, can you please explain what does disabling of the settings that I’ve shown above actually do?
