The following updates were released to the Kentik platform for AI / Insights during September 2026.
Features
- AI Advisor as your default landing page: You can now set AI Advisor as the page you land on after signing in.
- New AI Advisor sidebar and chat history: A persistent sidebar on the right side of the screen gives you chat history, recent chats, and AI settings. The sidebar is resizable, and chats auto-scroll until you scroll manually.
- Answer permalinks: Each AI Advisor answer comes with a permalink. Allows users who click a shared link to navigate to the answer.
- Inline SVG diagrams: AI Advisor now renders SVG diagrams inline, with zoom and export, instead of showing raw code. It adds to the existing Mermaid diagrams and can also generate charts from the data.
- Role-based access control for Runbooks: Runbooks are now subject to RBAC. Kentik-Managed Roles have been updated accordingly. Admins and super admins can create, edit, and delete runbooks. Standard members get read-only access by default, and AI Advisor respects these access levels.
- Fuzzy device matching: AI Advisor's device connections lookup now finds devices even when the name is slightly misspelled.
- Clearer sub-agent activity: Tool names now appear in readable form, and tool result panels are easier to follow.
- Longer sub-agent runs: The sub-agent tool timeout is now 10 minutes instead of 5, so AI Advisor can complete more complex multi-step analysis.
- AI model updates: Gemini 3.8 Flash and Claude Opus 5.5 were added, and the default models changed to Claude Sonnet 5 (advanced tier) and Gemini 3.5 Flash Lite (basic tier).
Bug Fixes
- Device config results in AI Advisor: AI Advisor now shows real device config results. You can compare revisions, search config content with context lines, and view the latest config.
- Markdown tables: Fixed broken table rendering in AI Advisor chat, Settings, Pathfinder reports, Insight Details and Boards panels.
- Command execution hangs: AI Advisor no longer hangs on commands that paginate output.
- Infinite loading in Custom Network Context: Fixed endless loading when the context is empty or a version is missing.
- Stale approvals: Fixed a race condition in the approval workflow that could resume a stale approval.
- Sub-agent replay: Fixed duplicate or incorrect information in conversation history after replaying sub-agent steps.
- Connectivity test errors: Failed connectivity tests now return the error text instead of breaking the agent step.
- Network metrics queries: Removed an invalid "none" aggregation option that caused query failures.
- Agents stopping early: Fixed AI agents stopping early when an OpenAI model returned commentary along with tool calls.