What on-call is, and why small teams dread it
Being on call means being the person who gets woken when something breaks. In a large company a rota shares it out. In a small one, the person on call is the person who wrote the feature, every night.
Most pages are not disasters. They are a crash affecting some users, a deploy that did not go out, a job that failed. Each one still costs someone their evening, and the fix is often a few lines.
What an on-call engineer does, step by step
| Step | A person | What AI can do |
|---|---|---|
| Notice | Gets paged | Hears about the error the moment it happens |
| Triage | Decides if it matters | Groups repeats into one problem |
| Diagnose | Reads logs and code | Reads the stack trace and the code, finds the cause |
| Fix | Writes a change | Writes the change and a test that proves it |
| Ship | Opens a pull request, waits for review | Opens the pull request with the evidence |
| Follow up | Checks it stayed fixed | Watches the change after it merges |
The diagnosis and the fix are where the hours go, and they are the steps a system can do while you sleep.
What Bugstack takes off the pager
Bugstack does autonomous software maintenance. For on-call, that means:
- A crash in production. Your app reports it. Bugstack reproduces it with a failing test, fixes it, and opens a pull request.
- A failed deploy on Render. It reads the build log, names the commit that broke it, and opens a pull request that fixes it or restores the last good commit.
- A red build on your main branch. It finds the cause and opens the fix.
In the morning there is a pull request to review. On a repository set to Autopilot, a proven fix merges without waiting for you.
What still needs a person
- Outages outside the code. A database that is down, a full disk, a provider having a bad day. Bugstack reports these and does not open a code change.
- Deciding to roll back. By default Bugstack opens a pull request that restores the last good commit, and a person merges it.
- Anything with no error. A page that is slow, or wrong, without failing.
- Talking to customers.
So this is not “no more on-call”. It is fewer pages that need you awake, and a head start on the ones that do.
Who this is for
- A team of two to ten with no rota and no site reliability engineer.
- A solo founder who is the whole on-call rota.
- A larger team that wants routine pages handled before a person is involved.
If you have a reliability team investigating incidents across many services, a tool built for investigation may fit that job better. Bugstack works on the code change and the upkeep between incidents. See Bugstack as an AI SRE.
Questions
Can AI replace on-call engineers?
It can handle the pages that come from a code fault and end in a code change. Infrastructure outages, rollback decisions and customer communication still need a person.
Will it fix things in production without asking?
Only if you set a repository to Autopilot. Every repository starts on Review, where Bugstack prepares the fix and a person merges it.
What does it need to work?
A GitHub repository, the Bugstack SDK in your app for crashes, and a Render connection for deploys. Failing builds need only the repository.