Skip to content
ONE3.ai

Engineering & Product · Digital Employee #100

AI Release Manager

Ship safely

Planned · Year 3 (W4)Autonomy ceiling A2Max risk without approval R3

Ship safely. AI Release Manager works for your product and engineering team under a named human owner, with its cost and results shown in the Investment & ROI Ledger.

What it does

Mission: Ship safely.

AI Release Manager works for your product and engineering team. Typical work: release notes, checklist, canary monitoring, rollback proposals. It works inside IT & Security — DevHub.

How it works under supervision

A named human owner is accountable for its work. Every hire starts in shadow mode (A1): it proposes, a person acts, and agreement is scored. In probation (A2) its drafts go out only after one-click approval until the probation criteria are met.

What it may do on its own

Its autonomy ceiling after probation is A2: it prepares each action as a draft and a person approves it with one click. Nothing it prepares takes effect without a person.

What needs approval

Its risk ceiling is R3. Once autonomy is earned, it may carry out external or customer-facing actions, such as sending a message, publishing or capped ad spend, without a separate approval. R4 Actions (payments, deletion, signing, payroll) always need a person. Production deploys always need human approval.

Cost and return in the Investment & ROI Ledger

Rent, AI usage, pass-through costs such as messages, and your team's supervision time are recorded as investment. Results are tracked against its KPIs (change failure rate, release lead time) and your previous process, shown as observed or assisted unless a reviewed experiment shows incremental return. Nothing is guaranteed.

Availability

This role is planned for wave W4, after Year 1, once the apps it uses are ready.

Skills & Actions

What it does

  • Release notes
  • Checklist
  • Canary monitoring
  • Rollback proposals

Tools

Apps it works in

  • IT & Security — DevHub

It uses only the apps and modules your workspace subscribes to, with the permissions you grant.

Frequently asked questions

Does AI Release Manager work without human oversight?

No. It has a named human owner and starts in shadow mode, then probation with one-click approvals. After probation its autonomy ceiling is A2: it prepares each action as a draft and a person approves it with one click.

What does AI Release Manager need approval for?

Its risk ceiling is R3. Anything irreversible goes to a person. R4 Actions such as payments, deletion, signing and payroll always need a person. Production deploys always need human approval.

How do I see what AI Release Manager costs and returns?

The Investment & ROI Ledger records its rent, AI usage, pass-through costs and your team's supervision time, and shows results against its KPIs compared with your previous process. Returns are labelled observed or assisted unless a reviewed experiment shows incremental value.

When will AI Release Manager be available?

It is planned for wave W4, after Year 1. One3 is built in waves, and a role is published only when all the apps it uses are ready.

Put the AI Release Manager on probation

Agree the goal, the budget and the approval rules. The Investment & ROI Ledger shows what it cost and what it returned.