AI Logs
Check the logs related to the workings of your AI Agents.
The AI Logs view lets you monitor and control everything related to the AI Agents enabled in your tenant.
The dashboard provides clear visual tools to measure agent efficiency, track session handling, and analyze errors so you can iterate on your agents with confidence.
You can also access logs related to a particular AI agent directly from the AI Agents view:
The available charts give you quick insight into AI Agents effectiveness:
Resolution Efficiency highlights total AI requests and sessions.
Session Severity Distribution summarizes session outcomes.
Session Error Trend shows error rates over the last four weeks.
The logs are grouped into requests, jobs, and sessions. For deeper insight into agent behavior, open the logs dedicated to a specific component or agent. Use search and the available filters in each view to find a specific log entry.
Each log entry is marked with the highest applicable severity: info, warning, or error. This means that if a given entry (request, job or session) contains an info and an error, it is marked with the error flag.
Switch to each tab to dig into the details.
Requests
The Requests view lists every request involving AI Agents. Open an entry to inspect Log Details, including the related metadata, the message and response, and the system log.
The Log Details view provides basic information about a specific entry:
Agent ID – identifier of the agent that handled the request
Session ID – identifier of the session this request belongs to
Request ID – identifier of this request
Duration – time the request took to complete
Created At – timestamp when the request was registered
Select Agent ID or Session ID to open the related agent or session.
The Communication panel shows the incoming message and the agent's response.
The Logs table lists system events for the request. Use the filters to find a specific entry. From here, you can also select Copy All Logs to copy the log entries to the clipboard.
Select Analyze Logs to start a conversation with the Agentic Log Analysis Assistant. The assistant lets you investigate the logs or look for a specific event.
If the assistant is not enabled yet, the first time you select Analyze Logs, you are prompted to enable the helper agent from the predefined template. Select Enable Helper Agent to create it. You can also enable it from the Predefined Agents list in the AI Agent Library. Once enabled, the agent remains available in the library.
After the helper is enabled, Analyze Logs opens a conversation modal. The assistant reviews the log messages, and you can ask follow-up questions directly in the conversation.
Jobs
The Jobs view captures system actions related to AI Agents, such as agent chat, import, or export operations. Open an entry to inspect Job Details and understand what succeeded, what failed, and why.
The Job Details view provides basic information about a specific entry:
Agent ID – The identifier of the agent related to the job. You can directly open the agent.
Job Type – The kind of system action, such as import, export, or agent chat.
Created At – Timestamp when the job was registered.
Status – The current state of the job, for example Finished.
The Communication panel shows the job message and the system response.
For import jobs, the Import Result section summarizes what was created or reused. It lists agents, tools, and MCP servers with their IDs and status, such as Disabled or To create. For more information about import outcomes, see AI Agent Import/Export.
Sessions
Use the Sessions view to inspect a specific AI session.
A session starts with the request when an agent is called, and it lasts through any collaboration with other agents or users invoked by this agent. For example, an incorrect invoice complaint triggers the Complaint Agent flow and the request starts a new session, which lasts through the whole agents flow:
Request -> Complaint Audit -> Complaint Categorization -> Complaint Agent -> Support Agent
Request (Slack or MS Teams message) -> Support Agent
Request (second Slack or MS Teams message) -> Support Agent
All steps above are within the same session. For custom agents using a communication tool such as MS Teams, subsequent requests from collaboration participants within the same conversation are also part of the same session.
Filter the logs by agent to focus on the relevant conversation. The details panel presents a linear timeline of every agent step, which gives a more business-oriented overview highlighting only the most important logs and all the warnings and errors. Select a card in the session to open the details view of the related log entry or entries.
Collaboration tool events
When an agent uses a communication tool (Slack or MS Teams), the session timeline includes collaboration-specific events:
User mapping failures – A configured participant could not be matched to a user in the communication platform or could not be added to the conversation. These entries are marked at warning or error severity.
Channel creation failures – The tool could not create or access the configured channel or group chat. These entries are marked at error severity.
Inbound messages – A reply from an identified user in the collaboration channel or group chat, forwarded to the agent that invoked the tool.
Unidentified users – An inbound message from a user who is not configured or recognized on the agentic platform. These entries are marked at warning severity.
Review these events to verify tool instance configuration and inbound handling settings. Conversation records are also stored on the communication tool instance and on the agent's Conversations tab. For configuration guidance, see Communication Tools Configuration.
Last updated
Was this helpful?