AI Assistant: Behaviour Modes

The Thrive AI Assistant (also known as “Kiki”) will soon be configurable to operate in different behaviour modes, giving your organisation control over how the assistant responds, which sources it uses, and how strictly it adheres to your company's data. This setting will be controlled by Super Admins and will apply globally across your platform.

Note:

For more information about the Thrive AI Assistant, see:

• AI Assistant: For Learners

• AI Assistant: Setup and Administration

Key Concepts

What is General Knowledge?

The Thrive AI Assistant is powered by large language models (LLMs), similar to mainstream tools like ChatGPT and Gemini. These models have a built-in understanding of the world, which means the AI Assistant will be able to answer general questions — for example, "What is leadership?" or "How does GDPR work?" — without needing that information to exist in your platform.

When we refer to "general knowledge" in this article, we mean this built-in capability, separate from your organisation's data.

What Counts As Your Content?

When the AI Assistant references "your content" or "platform content", this includes:

• Content inside Thrive: Learning content such as articles, videos, and e-learning, as well as events, assignments, and platform data.

• Trusted external sources: Whitelisted URLs configured via Trusted Web Search.

• Connected knowledge sources: Files and documents from integrations configured via Connector Search, such as Google Drive, SharePoint, OneDrive, and Dropbox.

  Note:

  For more information on searching across connected platforms, see Connector Search.

Together, these will form the AI Assistant's trusted knowledge ecosystem.

Behaviour Modes

Conversational Mode

Conversational mode will give the AI Assistant the most flexibility. It will respond freely and naturally — similar to a mainstream AI assistant — and will be able to engage in open-ended conversation on broader topics.

In this mode, the AI Assistant will:

• Use your platform content when relevant

• Draw on general knowledge freely

• Clearly distinguish between company knowledge and general knowledge

Best for: Engagement, exploration, and a flexible assistant experience.

Example prompts:

• "Explain what good leadership looks like"

• "Summarise this learning pathway"

• "Give me tips on time management"

Grounded with General Knowledge

This will be the recommended default. The AI Assistant will prioritise your organisation's content first and only use general knowledge when it can't find a relevant answer in your sources. When it does draw on general knowledge, it will clearly label the source.

In this mode, the AI Assistant will:

• Prefer your content first

• Use general knowledge only when necessary

• Maintain transparency about where answers come from

Best for: Balanced accuracy and flexibility — suitable for most organisations.

Example prompts:

• "What does our onboarding process look like?"

• "Explain GDPR in simple terms" (uses general knowledge if not covered in your content)

• "What training do we have on project management?"

Strict Sources Only

In this mode, the AI Assistant will only answer using your approved sources. If it can't find the answer, it will clearly say so. General knowledge will never be used.

In this mode, the AI Assistant will:

• Only draw on your approved content and connected sources

• Tell users when it doesn't have the information they need

• Never use general knowledge

Best for: Compliance, accuracy, and controlled environments — particularly regulated industries.

Example prompts:

• "What is our company's data retention policy?"

• "What compliance training is required for my role?"

• "What does our internal policy say about X?"

Configuring Behaviour Modes

1. In the Thrive menu, go to Configure > AI features.

2. Scroll down to Features > AI Assistant.

3. Under Response behaviour, select one of the following options:
   

   • Conversational: General chat mode — can discuss any topic freely

   • Grounded with general knowledge: Uses your content but can supplement with general knowledge

   • Strict sources only: Only responds using content from your platform

     Note:

     If you don’t configure a behaviour mode, the AI Assistant will default to this option. For more information, see Default Behaviour.

4. Click Save.

The selected behaviour mode is applied immediately and affects all users across your platform.

Default Behaviour

If no behaviour mode is configured, the AI Assistant will default to Strict sources only. This ensures consistency with the AI Assistant's original behaviour and avoids unintentionally changing how it responds for existing customers.

This means the AI Assistant will only respond using your approved sources — platform content, Connector Search, and Trusted Web Search. If it can't find an answer, it will clearly say so. General knowledge will not be used.

Important:

We recommend that admins explicitly select a behaviour mode to ensure it aligns with your organisation's needs. For more information, see Choosing the Right Mode.

Comparison

Choosing the Right Mode

• Want a flexible, ChatGPT-like experience? → Conversational

• Want reliable answers with flexibility? → Grounded with general knowledge

• Need strict compliance and source control? → Strict sources only