An AI employee for operations

Operations work is mostly noticing: that a process stalled, that a hand-off never happened, that something changed and nobody said. Dutta’s Operations employee watches continuously so the noticing is not a person’s job.

What it handles

Watches processes end to end. Follows work across the systems it passes through, rather than checking one queue. It sees the hand-off that was never picked up.

Fixes what stalls. Within the mandate you set, it restarts, re-routes and chases. Outside it, it raises the problem with the context already assembled.

Reports what changed. A written account of what moved, what stuck and what it did about it — not a dashboard you have to interpret.

Turns a recurring fix into a workflow. When the same intervention happens repeatedly, describe it once in plain language and it becomes a workflow that runs itself.

A shift, in order

ContinuouslyWatches the processes in its mandate for stalls, gaps and silent failures.
On a stallActs if the fix is within its permissions; otherwise raises it with the context attached.
DailyReports what moved, what stuck, and what it changed.
OngoingProposes workflows for the interventions it finds itself repeating.

What it can touch

An AI employee holds scoped credentials — read, write or send, per system — and nothing beyond its mandate. This role is typically granted:

  • Connected systems — read, to follow work across them
  • Named systems — write, only where you have granted it
  • Email — send, for the reports you asked for

Where a person stays in the loop

Dutta prepares, drafts and flags; a person approves anything with consequences. For this role that normally means:

  • Any fix outside the mandate you defined
  • Anything that changes a process rather than restarting one
  • Any action affecting a customer or a payment

Questions

What stops it acting where it should not?

It holds scoped credentials — read, write or send, per system — and nothing beyond its mandate. Permission is granted per system, not per employee.

How is this different from an automation?

An automation runs a script. An employee owns an outcome, asks when it is unsure, and reports what it did. That difference is the whole point of the product.

Can we replay what it did?

Yes. Every action is logged with inputs, sources, reasoning and output, and any run can be replayed in full.

Hire this role

Pricing is per AI employee on shift; humans are unlimited on every plan. See pricing or tell us the process that eats your team’s week.

Other roles

finance operations · customer support · recruiting · growth · Custom roles