[FAQ] io.Assist permission management

io.Intelligence
1

Question

Can we set different permission levels for different user groups, and what happens when co-pilots encounter ambiguous situations or conflicting data across systems?

Answer

Yes, setting up different permission levels is a key part of the architecture and design of io.Assist.

  1. User Awareness: io.Assist is intended to be aware of a user’s role or group permissions, leveraging whatever existing authentication or entitlement system the organization already has in place.

  2. Dynamic Prompts: The prompts that appear in the co-pilot interface are not necessarily universal. They can be dynamically curated and provided based on the user’s specific role.

  3. Tool Control: This dynamic curation is also true for the tools available to the co-pilot. Administrators have full control over the prompt libraries and the tools that are available.

  4. Judicious Provisioning: We recommend limiting the number of tools provided to an AI system to the ones it needs to achieve it’s goal. Providing every tool possible just because we can is in general a bad idea . Be cautious in what you provide to end-users and what actions you are comfortable with the AI taking on the user’s behalf.

  5. External Access: Organizations can decide whether they want co-pilots to have access to web search (external sources) or if they should be limited strictly to internal knowledge centers and repositories of information. These are enterprise decisions based on the organization’s risk tolerance.

  6. Fine-Grained Controls: The permissions can be very fine-grained, and the system includes controls in the background. For instance, administrators (not end-users) should set parameters like LLM temperature or token usage.

2 Likes