Today we're announcing Lumos AI — a deeply integrated AI layer that works across your entire workspace to reduce toil, surface insights, and help your team focus on what matters.
This isn't a chatbot bolted onto a sidebar. Lumos AI is woven into the core product: issue creation, triage, summarization, cycle planning, and more.
The Problem: Backlog Rot
Every engineering team knows the feeling. The backlog grows. Issues lose context. Nobody is sure which bugs are still reproducible. Priorities drift. The backlog stops being a planning tool and becomes a guilt museum.
AI triage exists to stop backlog rot before it starts.
What Lumos AI Can Do
Auto-triage incoming issues
When a new issue is created — from a customer report, a Slack message, or a GitHub issue — Lumos AI reads it and:
- Labels it with the right type (bug, feature, improvement, security)
- Assigns a priority based on impact signals and similar past issues
- Routes it to the right team or individual
- Flags duplicates so you're not solving the same problem twice
All of this happens in seconds, before a human ever looks at it.
Summarize long threads
Engineering discussions get messy. A GitHub PR thread with 60 comments, a Slack thread with 40 messages, a document with weeks of edits — Lumos AI reads them all and generates a structured summary with the decision, open questions, and next steps.
Write better issues
Start typing an issue title and Lumos AI suggests a full description, acceptance criteria, and relevant labels based on your team's historical patterns. Junior engineers create issues that look like they were written by seniors.
Identify blockers early
Lumos AI monitors your active cycles and flags issues that are showing signs of becoming blockers — no updates for 3 days, dependencies not resolved, assignee at capacity. It alerts you before the slip happens.
Cycle planning assistance
At the start of a new cycle, AI surfaces a suggested set of issues based on current priorities, team capacity, and what historically gets completed. You click accept or make adjustments — no more staring at a blank planning board.
How We Built It
Lumos AI uses a combination of models depending on the task. For triage and classification, we use fast, small models that operate with sub-100ms latency. For summarization and generation, we use larger models with richer context windows.
All AI features can be turned off per workspace. Your data is never used to train any external models.
Getting Started
Lumos AI is available on all paid plans starting today. Head to Settings → AI to configure triage rules, routing logic, and which AI features are active for your workspace.
We're just getting started. The roadmap includes AI-generated retrospective reports, predictive velocity, and natural language issue search. Stay tuned.