What you will learn
- Navigate the unified omnichannel conversation view in raia Copilot.
- Use the three primary Copilot modes: Monitor, Debug, and Intervene.
- Understand how to use Admin Mode to diagnose agent behavior.
- Execute a seamless human takeover during a live conversation.
The Three Ways to Use Copilot
raia Copilot is not just a dashboard — it is your primary workspace for collaborating with AI agents. As an Agent User, your role is to oversee the agent's work, step in when human judgment is required, and provide the feedback that makes the agent smarter.
Mode 01
Monitor Mode
Your default state. Watch live conversations unfold across Live Chat, SMS, Email, and Voice in a single unified feed. You look for trends, monitor sentiment, and make sure the agent is handling routine inquiries smoothly.
Mode 02
Debug Mode
When an agent gives an unexpected answer, switch to Debug (by enabling Admin Mode). Look under the hood to see what the agent was thinking, what documents it retrieved, and why it produced that response.
Mode 03
Intervene Mode
When a conversation needs empathy, complex negotiation, or falls outside the agent's guardrails, click Take Over. The agent pauses and you step into the conversation seamlessly.
What Admin Mode Shows You
To oversee an agent effectively, you need to understand why it says what it says. Admin Mode transforms the agent from a black box into a transparent, debuggable system.
When you toggle Admin Mode on, a side panel opens next to the conversation thread displaying four critical pieces of information for every agent response:
01
Knowledge Retrieved
Exactly which chunks of text the agent pulled from the Vector Store, including their relevance scores.
02
Reasoning
A plain-text summary of how the agent interpreted the user's intent.
03
Skills Used
A log of any external tools the agent triggered (e.g., calling an API or checking a database).
04
Confidence
The agent's internal confidence score for the response.
If the agent gives a bad answer, Admin Mode tells you exactly whether it was a failure of knowledge (it retrieved the wrong document) or a failure of instruction (it retrieved the right document but applied the wrong rule).
The Human-in-the-Loop Decision
The most important decision you make as an Agent User is knowing when to let the agent handle it, and when to step in.
Seamless Takeover
When you click Take Over, the transition is invisible to the end user. From their perspective the "Support Agent" simply continues the conversation — but with your human nuance. Once you have resolved the complex issue, you have two choices:
Close the conversation
Use when the interaction is fully complete and no further follow-up is needed.
Hand back to agent
Use when the user has more routine questions the AI can handle, freeing you up to assist someone else.
Frequently Asked Questions
Does the user know when I take over a conversation?
No. The transition is seamless. However, depending on your company's policies, you may choose to introduce yourself (e.g., "Hi, this is Sarah stepping in to help with this complex issue").
Can I see conversations from all channels in Copilot?
Yes. Whether the user is texting via SMS, emailing, or using the website Live Chat, all conversations appear in the same chronological feed in Copilot.
If I fix an answer during a takeover, does the agent learn from it automatically?
Taking over solves the immediate customer issue, but to make the agent permanently smarter you must also use the Feedback and Scoring tools, covered in the next module.