Back to blog
ArticleBlog / Workflow Automation
Jun 22, 20264 min readWorkflow Automation

Approval process automation: the two minutes that matter

Automating an approval process is not removing the approver. It is putting the exact change in front of them, then recording the decision. How Tensor does it.

Written by Tensor AutonomousThe Tensor Autonomous team builds approved AI Action and workflow automation systems for service businesses.

Every approval has two parts: the ten minutes of finding out what is being asked, and the two minutes of deciding. Automation that removes the two minutes is dangerous. Automation that removes the ten is the whole point.

This is how Tensor Autonomous automates the approval process around its own work: everything before the decision is done for you, the decision stays yours, and the record is kept without anyone writing it.

The ten minutes, done before you arrive

When one of Tensor's responsibilities reaches a change, you do not get a notification that says "a change needs approval". You get the change:

  • for a booking, the event with its title, time and calendar;
  • for a log entry, the row with every cell filled from the request, and the request itself;
  • for a reply, the full text, addressed.

Nothing to look up. The context that used to take ten minutes is on the screen with the question.

The two minutes, kept

You approve, edit or decline. Tensor shows you each change before it makes it, drafts included; nothing it drafts is sent until you send it. That is the default for every new responsibility.

The one thing automation should never do is make this part feel like a formality. Tensor shows the real change, not a summary of it, so the two minutes are spent on the decision rather than on trust.

Making the two minutes rarer, not shorter

The honest way to reduce approvals is to decide, once, which ones you do not need to see. In Tensor that is a limit per responsibility: which kinds of change may run without asking (adding a row, creating a draft, adding an event), a daily cap, a days-and-hours window, and for each write an amount ceiling and a limit on recipients. Tensor may book without asking on weekday mornings, up to three a day, say; it may add rows to the maintenance log without asking. Within the limits the work runs; at the edge it stops and asks. Sending email, payments and refunds, and moving or cancelling a calendar event always ask; limits can't cover them.

Limits are set after you have watched the routine run right, not before. Approval first, limits later, is the order that keeps the limits honest.

The record, written by the process

Every run leaves a receipt: what was asked, what Tensor read, what it proposed, who approved it and when (or which limit it ran under), what it wrote, and the read-back confirming the write matched. Nobody fills in a form. When the question "who approved that?" comes up, the answer already exists.

Next to each write, Undo removes what Tensor created after you confirm what will be removed, and a responsibility can be rolled back to an earlier version. Sent email cannot be unsent, which is why every send waits for your OK; limits can't cover sending.

What this is not

It is not approval-routing software for requests between people: multi-step sign-offs, purchase requests, HR forms. Those tools solve a different problem, and a business with layers of sign-off may need one. Tensor automates the approval of its own work, which in a small business is the approval that happens forty times a week.

A quick way to see it

The live demo is one approval process end to end: you give Tensor a sample task in your words, it reads the sample records, puts the exact change in front of you, waits, makes the change on your OK, reads it back and shows the receipt. Two minutes, and you will know whether this is the kind of approval you want to automate.

Where to start

See it in a demo

The exact change, your two minutes, the receipt.

Try it live

Tagged in#approval process automation#workflow automation#AI Actions