Even though scripted agents are blocked with a checkbox on the region estate settings. There should be a second checkbox that allows ConvAI bots in even when the prevent scripted agents checkbox is set. That way region owners can keep scripted agents out but allow ConvAI bots to visit.

Update: Alternatively, a "Capabilities" concept could be provided for scripted agents. This would be similar to how phone apps have to declare which features of the phone they need access to, and then when users install an app they can decide whether they want to continue installing and allow those capabilities. In other words, you'd still only have 1 checkbox on the estate management for allowing or disallowing scripted agents. But, then there would be button that opens a separate dialog if scripted agents are allowed, that displays a series of checkboxes for the region estate manager to decide which kind of bots they would allow. By default all these checkboxes would be enabled until the manager unchecks them and hits apply.

Examples of capabilities; tracking users, conversational (convai), land/group moderation (e.g. smartbots), etc. I guess the only concern would be, bot owners could lie and declare the wrong capabilities and there'd be no way of LL to know. Whereas mobile device companies are able to scan the code during compile-time for usage of certain API. I'm not sure how to really solve that, unless LL has a TPV-like gatekeep registration process where only certain companies are allowed a specific capability, e.g. smartbots are known to only do land/group moderation and are trusted so they'd automatically get access to providing the moderation capability, and convai bots LL is providing are already trusted to only provide conversational responses when prompted so they'd receive that capability.

Also, possibly LL could require that the code for their scripted agent be on github and presented for review, to verify capabilities. If a company has a proprietary scripted agent and refuses to allow reviews of the code, then there could be a Capability called "Unreviewed / Proprietary" and that category is understood to mean that LL has no way of knowing what the scripted agent is up to, but at least the region managers are able to decide if they want to risk allowing those or optionally then disallowing them but allowing others that have declared their capabilities and were code reviewed. Regardless how it is implemented ultimately, having this as a separate dialog and set of checkboxes would provide for more types of bots to be whitelisted based on capability. Whereas my previous suggestion above was a specific checkbox only for convai and didn't take into account the vast variety of other types of scripted agents available.