The following updates were released to the Kentik platform for AI / Insights during October 2026.
Features
Streaming responses and Stop button: AI Advisor answers appear as they are generated, and the view follows the response as it grows. Claude models also show a collapsible "Reasoning" summary above the chat input. While an answer runs, Send becomes Stop in the main chat and the copilot.
Infoblox integration: A new Infoblox tab in Settings > Kentik AI (admin access) connects AI Advisor to Infoblox, with support for multiple realms. This integration allows customers of both Infoblox UDDI and Kentik to ask AI Advisor questions pertaining to their Infoblox setup.
New models: Claude Sonnet 5.5 is the default for the advanced tier and is generally available. Claude Haiku 5.5 is added to the basic tier.
AI Advisor sidebar: Sections now match Data Explorer (Sessions, Recent, AI Settings). The sidebar moves to the right, with a Hide/Show Sidebar button.
Journeys v1 retired: The old Journeys page is removed, and /v4/core/journeys redirects. Journeys predates AI Advisor and is now fully replaced by it.
Runbook permissions: Chat viewers without runbook access see "You don't have permission to view runbooks" instead of runbook results. This comes with full RBAC support for runbooks.
Faster failure for slow NMS queries: The NMS tool stops after 3 minutes and suggests how to narrow the query.
Runbook approval: Runbook steps that run sensitive commands ask for approval and show a preview (human in the Loop/Elicitation prompts.
Bug Fixes
Cut-off answers: Final answers are no longer cut at 4,000 characters, which affected answers with inline SVGs.
Sub-agent errors: Sub-agents no longer fail with a "tool_choice not supported" error on newer Claude models.
Thinking indicator: It now stays visible while sub-agents run, and tool-calling steps no longer stay stuck as "Processing".
Long chats in the copilot: Opening a long chat from history now scrolls correctly.
Export and summary order: Exports, summaries, and load order now skip sub-agent steps correctly.
Invalid tool arguments: Tools now return an error for unknown dimensions or invalid IDs, instead of returning wrong or empty results.
AI Site Search: A search with no matching pages no longer returns a 500 error.
AI settings permissions: AI settings endpoints now respect user RBAC permissions.