FIELD GUIDE

The first workflow sets the operating standard.

Teams often begin with a broad ambition: improve support, automate sales, or make operations faster. Those are worthwhile goals, but they are too wide to become a dependable first job. A strong AI employee starts with one workflow that has a clear trigger, a visible outcome, and a human owner.

The right first workflow is not necessarily the largest opportunity. It is the smallest useful loop that can produce trustworthy evidence about knowledge, actions, handoffs, and customer experience.

Look for repetition with real value

Start with work that happens frequently and follows a recognisable pattern. Good candidates include qualifying an inbound enquiry, answering questions from an approved knowledge base, collecting intake information, preparing a meeting brief, or routing a request to the right team.

The workflow should matter enough to measure, but remain contained enough that mistakes can be reviewed and corrected without creating unacceptable risk.

Write the job as an operating contract

Before configuring prompts or choosing a model, describe the job in plain language. Name the trigger, the information the employee may use, the actions it may take, the situations it must not handle, and the person who owns the outcome.

  • Trigger: what starts the workflow?
  • Outcome: what useful state should exist when it ends?
  • Knowledge: which approved sources may guide the response?
  • Actions: which tools are allowed, and which require approval?
  • Handoff: when should a person take over?
  • Evidence: what will the team review each week?

Choose a channel that matches the job

A website conversation, phone call, WhatsApp exchange, and internal Slack request each create different expectations. Select the first channel because it fits the workflow, not because it is the easiest place to add a widget. The same role can expand later once its boundaries are proven.

A useful pilot is not a demonstration of maximum autonomy. It is evidence that one job can be completed reliably inside a clear boundary.

Measure the complete loop

Track completion, accuracy, response time, handoff quality, and the amount of human correction required. These measures reveal whether the workflow is genuinely helping the team. When the loop is stable, expand deliberately: add knowledge, another channel, or one more approved action at a time.