Skip to content
Last9
Book demo

AI Assistant

Investigate errors, latency, and alerts in a Last9 AI chat conversation, then open the underlying telemetry in the dashboard.

AI Assistant is the chat surface for Last9 AI in the dashboard. Ask about a service and time window, review the evidence behind each answer, and open the underlying telemetry in Last9.

Start with a specific service, environment, and time window:

Investigate HTTP 5xx errors for checkout-api in production over the last 30 minutes. Confirm the service and data source, identify the affected routes, and include links to the supporting evidence. Separate confirmed findings from possible causes.

Open as draft in AI Assistant

Replace checkout-api with your service name. For the full investigation method, refer to Investigate an incident with AI.

Prerequisites

  • AI access enabled for your Last9 organization. See Enable Last9 AI.
  • Relevant logs, metrics, or traces flowing into Last9.

Start a conversation

  1. Open AI Assistant in the Last9 dashboard.

  2. Ask about a service and time window. Include an environment, alert, request ID, trace ID, or error message when available.

  3. Review the answer and open its evidence links. Check that the query, filters, and time range match your question.

  4. Ask a follow-up in the same conversation. Focus on a service or request, compare with successful requests, or ask which evidence is still missing.

AI Assistant welcome screen

Work in a conversation

Once a conversation is open, use it to follow the evidence, check an alert, and keep the signals you verified.

Investigate errors and latency

Ask the assistant to inspect exceptions, log patterns, service performance, and traces. Useful follow-ups connect an observed symptom to a specific request or dependency:

For the failing request identified above, inspect the related trace and logs. Which operation failed, and what evidence supports the explanation?

Open as draft in AI Assistant

Compare slow and successful requests for the selected route in the same window. Which operations or attributes differ?

Open as draft in AI Assistant

The assistant can only investigate data available to its tools. Missing instrumentation, restricted access, and incomplete results can leave the cause unresolved.

Review alerts and changes

Use an alert as an investigation starting point. Ask which condition fired, which services are affected, and whether available change events overlap with the incident.

Investigate the alert for the window in its message. Check the supporting telemetry and recent change events. Explain what is confirmed and what needs further investigation.

Open as draft in AI Assistant

Author an alert rule

Ask the assistant to draft an alert rule for a verified signal. Last9 AI validates the expression, creates the alert rule disabled as a draft, and asks for approval before enabling it. Alert rule authoring depends on your organization’s AI settings and on whether the connected Last9 MCP session serves the alert write tools.

Draft a p95 latency alert for checkout-api in production on the existing alert group. Validate the expression against the group's KPIs, create the rule disabled, and wait for my approval before enabling it.

Open as draft in AI Assistant

Review the proposed threshold, KPI, and notification routing before you approve promotion. For the MCP tool flow behind this workflow, see Last9 MCP → Alert management.

Ask for a trend chart

When your question asks how a metric or log count changed over time, the assistant can attach a grounded trend chart built from in-scope range reads. Ask for the series, window, and comparison you care about:

Chart checkout-api request rate and error rate in production for the last 6 hours in 5-minute buckets. Call out any sustained increase.

Open as draft in AI Assistant

In Slack, trend charts appear as tables in the thread.

Keep the environment on follow-ups

After you confirm a service and environment, later turns in the same conversation reuse that choice unless you change it. Correct the environment explicitly when a follow-up should switch scopes.

Responses can include links to the underlying data in Last9. Use them to inspect the query, continue in the explorer, and share evidence with a teammate.

Use the linked query as the starting point for additional filters or a more detailed investigation.

Create a dashboard from an investigation

Ask the assistant to create a dashboard for the signals you want to keep watching:

Create a dashboard for checkout-api using the request rate, error rate, and latency metrics we verified in this investigation. Preview the panels before saving, then wait for my approval.

Open as draft in AI Assistant

The assistant can propose creating or updating saved dashboards, preview queries before save, and apply targeted panel edits. Review the operation and its arguments before approving the change. For an existing dashboard, provide its URL and ask for an update.

Chat history

Use chat history to return to an investigation and continue asking questions. Start a new chat for an unrelated incident so its details stay separate.

When reopening an older conversation, ask for a new query window before relying on its recovery status. Previous answers describe the data queried at that time.

Ask mode in Logs and Traces

For a focused query in Logs Explorer or Traces Explorer, use the Ask tab to describe the records you want to find:

Show errors for checkout-api in production over the last 30 minutes.

Review the generated filters and matching results. Use AI Assistant when you need a conversation that follows evidence across signals.

Ask mode in Logs Explorer


Troubleshooting

  • AI Assistant is unavailable

    Ask an organization admin to check Enable Last9 AI in AI Settings.

  • The answer covers the wrong service

    Confirm the service name, environment, and data source, then correct them in a follow-up.

  • No data is returned

    Check the time zone, time window, filters, and whether the relevant signal is ingested.

  • The analysis is incomplete

    Ask which queries failed or returned partial results. Resolve those gaps before relying on the conclusion.

  • The assistant cannot query Last9 tools

    Ask which MCP tools were unavailable or refused. Check Enable Last9 AI in AI Settings, then retry with a narrower service and window.

  • Alert authoring is refused

    Confirm that AI alert authoring is enabled for your organization and that the session can use Last9 MCP alert write tools. Ask for a draft recommendation without saving if write tools are unavailable.

  • A saved answer no longer matches current conditions

    Ask the assistant to query a new explicit window.

Please get in touch with us on Discord or Email if you have any questions.