AI Fellows

An AI Fellow is a configurable conversational AI owned by the company. Users open a session and interact with the AI Fellow through voice, similar to a video call. Each AI Fellow has its own avatar, voice, name, language, instructions, knowledge documents, and Connectors.

AI Fellows can be deployed standalone via a public link, or assigned to a Learning Path to guide users through structured sessions.

When to use

Use an AI Fellow when a single live conversation is required, rather than a multi-session program. Common use cases include coaching, simulations, technical onboarding, compliance walkthroughs, leadership practice, and public demos accessible from a link.

For ordered sessions, objectives, cohort analytics, and per-user progress, use a Learning Path instead.

Setup process

Start with Create an AI Fellow. The setup flow covers avatar, name, language, starting point, and purpose. When the goal is a role-play exercise, start with AI Simulations instead. The purpose is the most important configuration field for standard AI Fellows, so review the Define the purpose guidelines.

After the basics are set, add reference material in Knowledge Base, enable Connectors, and review the AI Fellow's Visibility. Private AI Fellows also expose a Manage Users page for granting access.

To distribute the AI Fellow, use Deployment options and the Integration Guide for the standalone link, embed code, language options, and external user tracking.

Functional overview

An AI Fellow can:

• Hold a live, voice-first conversation in real time.
• Run as a text conversation instead of voice when an admin has enabled chat mode. The user types, the AI Fellow replies in text, and the conversation is saved like a voice conversation.
• Answer questions grounded in the documents uploaded to its Knowledge Base.
• Use a file that the user drags into the current session as temporary context for that conversation. See Files sent during sessions.
• Run a realistic simulation, for example a skeptical CFO, a difficult customer, or a regulator asking pointed questions.
• Walk a user through a shared screen and comment on what is shown.
• Collect structured information during the conversation and submit it through a Connector to one of the customer's systems.
• Resume a previous conversation when the user chooses to continue one.
• Open with an optional Info screen shown whenever the user opens the AI Fellow page. This is useful when the user needs a short briefing before the conversation starts.

A live session sounds like a phone or video call. The user speaks, the AI Fellow responds, and the exchange happens in real time. A session can also include camera, screen, and text sharing. Camera and text sharing are controlled by the user. Screen sharing appears when it has been enabled for that AI Fellow or Learning Path.

A single live session is capped at 45 minutes. When the cap is reached, the conversation ends and the user can start a new one if they want to continue.

Limitations and intended use

• AI Fellows are not general-purpose chatbots. Even when chat mode is enabled, they are designed for structured conversations that can trigger approved actions.
• AI Fellows are not a substitute for a human reviewer. Final judgment must remain with a person.
• AI Fellows are not autonomous decision-makers. Connector executions are scoped to the conversation.
• AI Fellows are not a guaranteed source of truth. As with any AI system, output may be incorrect; periodic transcript review is required.
• AI Fellows are not authorized to provide legal, medical, financial, or compliance advice.

Monitoring and optimization

Use Available actions as the reference for operations available from the list and monitor views. Once users start interacting with the AI Fellow, review Conversations and analytics and Conversation management to review usage, read transcripts, and delete conversation contents when required. Use Edit history to see which admins changed the AI Fellow and what changed.

The AI Fellows list can be filtered by status, type, usage in the selected period, access state, and creator. It can also be sorted by title, creation date, conversations, or access count.

Configurable fields: