Continuous Learning
Autoheal improves with every investigation. As your team works through incidents, the platform captures what was learned and applies it automatically next time. Investigations get faster, and more issues are caught before they page anyone.
This learning happens on its own. You don't have to write anything up for it to take effect (though you can author context yourself too).
What Autoheal learns from
Learning is grounded in real signals from your investigations, not generic best practices:
- Accepted root causes. When you confirm a root cause, that diagnosis becomes reusable knowledge for similar incidents.
- Guidance during investigations. When you steer, correct, or add context mid-investigation, Autoheal incorporates that direction.
- Past discoveries. When an investigation takes real effort to uncover a non-obvious cause, that finding is kept so the next occurrence is quick to resolve.
How that learning pays off
Captured learning is applied through three mechanisms:
| Mechanism | What it does | Impact |
|---|---|---|
| Related investigations | Surfaces similar past investigations as a new one begins, so prior diagnosis and resolution paths are reused instead of rediscovered | Faster time to understand |
| Memories | Reusable learnings from past incidents that are recalled and applied automatically in future investigations | Reduces MTTR (mean time to resolve) |
| Proactive actions | Grounded improvements surfaced from past investigations — tune alerts, improve observability, fix code, improve testing | Reduces MTTD (mean time to detect), or prevents the incident entirely |
Related investigations
When an investigation starts, Autoheal looks for past investigations that resemble it and surfaces them in context. Instead of starting from a blank slate, the investigation builds on what already worked: the data sources that mattered, the hypotheses that panned out, and the resolution that fixed it.
Memories — reduce MTTR
Memories are reusable learnings drawn from past incidents. Once captured, they're recalled automatically when a relevant incident recurs, so a problem your team has seen before is resolved quickly rather than re-investigated from scratch. This lowers mean time to resolve (MTTR). Memories are learned automatically, and you review and approve what's captured. See Memories for details and the full lifecycle.
Proactive actions — reduce MTTD or prevent
Some learnings are best applied before the next incident. Autoheal surfaces proactive actions: concrete improvements like tuning a noisy or missing alert, closing an observability gap, fixing a recurring root cause in code, or adding a test. Acting on them helps catch issues sooner (lower mean time to detect, MTTD) or stops them from happening at all.
You stay in control
Autoheal suggests; your team decides. Learnings are surfaced for review, and proactive actions are approved before anything changes in your systems. Nothing is applied to your systems automatically.