HANDORU
03 / Safety and control

AI can talk freely. It cannot commit freely.

Every real-world action should pass through explicit permissions, customer confirmation, system verification, and a durable record.

See the commitment model ↓
Customer commitments

Prepare. Confirm. Execute. Verify. Record.

This is the product principle. The exact actions available in a pilot depend on approved integrations and policies.

Connected mode / illustrative flowCustomer commitment
IN
Customer request

Can I book a dental hygiene appointment tomorrow afternoon?

Current operation

Tomorrow afternoon, dental hygiene

In progress
Before and after connection

First it learns. Then it can act.

Connecting a business tool changes what the agent is allowed to promise.

Before tools are connected

Answers and prepares.

Customer

“Can I come tomorrow afternoon?”

Handoru

“I can explain the service and collect your preferred time. I cannot confirm the booking yet.”

  • Answers from approved information
  • Collects the customer’s request
  • Does not claim an action is complete
After the calendar is connected

Checks and completes.

Customer

“Can I come tomorrow afternoon?”

Handoru

“I checked the live calendar. Friday at 15:30 is available. Shall I book it?”

  • Checks live availability
  • Waits for customer confirmation
  • Books and records the verified result
When a person is needed

Handoru brings in your team.

If a request needs a human decision, Handoru pauses and asks the right teammate for help.

01 / PAUSE

It knows when to stop.

Handoru does not guess when a request is sensitive, unusual, or outside its instructions.

02 / EXPLAIN

Your teammate gets the full story.

They see what the customer wants, what has already happened, and what decision is needed.

03 / CONTINUE

The customer does not start over.

Your teammate joins the same conversation and helps from exactly where Handoru stopped.

Decide what Handoru can do and when your team steps in.

Join waitlist →