Skip to content

Tentacles V1 Workflow

Purpose

This document makes the first Tentacles workflow concrete.

First Workflow

The first live workflow is for TV Mount Australia.

It should handle:

  • customer enquiries
  • basic qualification
  • booking-adjacent requests
  • operator escalation when the system is uncertain or the action should be reviewed

Primary Customer Paths

WhatsApp

  1. customer sends a WhatsApp message
  2. workflow identifies tenant and request type
  3. system attempts knowledge-assisted response
  4. if confidence is high and the reply is safe, respond
  5. if confidence is low or review is needed:
  6. send holding response
  7. create or update internal review queue
  8. notify operator

Email

  1. customer sends an email
  2. workflow identifies tenant and request type
  3. system attempts knowledge-assisted response
  4. if confidence is high and the reply is safe, respond
  5. if confidence is low or review is needed:
  6. send holding response if appropriate
  7. create or update internal review queue
  8. notify operator

Operator Workflow

The first operator model is:

  • internal review queue
  • human notified by email or WhatsApp when queue is not empty
  • customer receives a holding response while waiting

The operator must be able to:

  • review the request
  • approve or replace the response
  • trigger next step for booking-related handling

Booking Boundary

For the TV Mount Australia pilot, booking integration is expected through Google Calendar.

The first booking-capable workflow does not need full autonomous scheduling. It only needs a controlled path that moves from enquiry to human-reviewed booking action.

V1 Success Criteria

V1 is successful when it can:

  • receive real inbound messages
  • answer from approved knowledge
  • create a real queue when needed
  • notify a human operator
  • support a controlled booking-related path
  • demonstrate one useful end-to-end pilot for TV Mount Australia

Explicit Non-Goals For V1

  • full voice integration
  • broad web chat rollout
  • self-service tenant onboarding
  • perfect automation without operator review
  • one full stack per tenant