Context
An AI assistant working inside operational workflows at a business in regulated finance, where actions such as submitting or closing carry real consequences.
The problem
In a regulated business, an AI that can take actions needs clear limits. Decisions to submit, close, or finalize anything with consequences had to stay with an accountable person.
Constraints
- Consequential actions need an accountable person behind them, not the AI.
- Approval might not come for days, so paused work had to resume exactly where it stopped.
- Every decision, allowed or denied, had to be on the record.
Approach
Every action the AI proposes passes through a policy gate. If no rule explicitly permits it, it is denied, and every decision is logged. Anything with real consequences does not run at all: the workflow pauses, shows what is waiting, and only a person’s approval starts it again, picking up where it stopped, even days later.
Key engineering decisions
Deny by default
The gate denies anything that is not explicitly permitted (it “fails closed”), so an action nobody anticipated is refused rather than allowed by omission.
Pause and resume on approval
Consequential steps save the workflow’s progress and wait. Approval resumes it from that point instead of starting over, so a long wait loses no work.
Log every decision
Allowed and denied actions are both recorded, leaving a clear record of who decided what, and when.
Result
The AI proposes; a person approves. Nothing important happens on the AI’s say-so, and there is a clear record of who decided what, and when.