HIATT CO / AI TRAININGAll programs ↗

A POSSIBLE SESSION FOR YOUR TEAM

Sample course outline — adapted after a planning conversation.

AI Business Systems: design a controlled request workflow

Map a familiar process, keep fixed decisions in clear rules, and test where AI can help draft text under human review.

PROPOSED TIME120 minutes total

Proposed: a facilitated 120-minute working session.

PROPOSED TOOLS

Proposed: laptop, notes or a drawing tool and a team-approved text AI tool. Test cases run manually; connecting live systems is outside this sample.

This example is a starting point. Scope, timing, tools and deliverables are agreed together before a session.

Download standalone HTML ↓

WHO IT SERVES

Is this your next step?

Owners and operations leads who understand a repeatable process and already use AI for drafts.

Before the session

  • Ability to describe a process from intake to completion.
  • Experience checking AI-written text.
  • Awareness of who owns approvals in the process.

PRACTICE & TAKEAWAYS

Work you can try.

What you will practice

  • Map inputs, rules, AI assistance and human approvals for one process.
  • Write explicit stop conditions for missing or conflicting information.
  • Run normal and exception cases and record whether the workflow follows its rules.

What you can leave with

  • A one-page workflow map.
  • A rule table and approval checklist.
  • Three manual test records and a list of questions for any future implementation.

PROPOSED AGENDA / 120 MINUTES

From first example to review.

Open each section for the exercise, output and review check. All practice uses fictional or anonymized material.

01Map the current process25 min

Purpose. Start with a process people can already explain.

Learning outcome. Identify the intake, decision owner and final action.

Explain or demonstrate. Trace a fictional service request from receipt through review to a customer reply.

Hands-on practice. Map Meadow Repairs intake: request ID, customer message and service area. Service areas are North and Central. Staff confirm appointment availability. An office lead approves every outbound reply.

Your output. A process map with named human approval and sending steps.

Review check. Map ends with a person approving and sending; no appointment is confirmed without an availability check.

Prerequisites. Familiarity with a repeatable business process.

Proposed time: 25 minutes, including the explanation, practice and review.

02Write the rules before the prompt25 min

Purpose. Use simple rules for fixed conditions.

Learning outcome. Define a route for valid, missing and unsupported service areas.

Explain or demonstrate. Show a rule table: North or Central → prepare draft; missing area → request clarification draft; other area → office-lead review.

Hands-on practice. Write the rule table and identify which steps need no AI. Add a stop condition if sources conflict or an area is ambiguous.

Your output. A rule table with inputs, route and escalation owner.

Review check. All routes preserve human approval; ambiguity and conflicting facts stop normal processing.

Prerequisites. Current-process map.

Proposed time: 25 minutes, including the explanation, practice and review.

03Break5 min

Purpose. Pause before the next practice round.

Learning outcome. Rest time; no learning outcome assessed.

Explain or demonstrate. No demonstration during the break.

Hands-on practice. Step away from the exercise.

Your output. No participant output required.

Review check. No assessment during the break.

Prerequisites. None.

Proposed time: 5 minutes, including the explanation, practice and review.

04Draft within the boundary30 min

Purpose. Give AI a limited writing task after routing.

Learning outcome. Create a reply that follows a selected route without making a booking.

Explain or demonstrate. Provide a route explicitly in the prompt rather than asking AI to invent policy.

Hands-on practice. Use the representative North-area request. Manually select the route, then ask AI for a reply and check every factual claim.

Your output. A reply draft with request ID and approval status.

Review check. Draft acknowledges the request, leaves availability unconfirmed and is marked for office-lead review.

Prerequisites. Agreed rule table.

Proposed time: 30 minutes, including the explanation, practice and review.

05Test the normal and exception paths35 min

Purpose. Check the process before considering connections.

Learning outcome. Run three cases and document route, draft and human decision.

Explain or demonstrate. Show a test record with expected route, observed result and pass or revise.

Hands-on practice. Run R-21: North area, asks for Friday; R-22: area missing, asks for Tuesday; R-23: South area, asks for Monday. No availability is supplied. Correct any draft that promises an appointment or skips approval.

Your output. Three test records and a revised workflow map.

Review check. R-21 prepares a draft; R-22 requests area clarification; R-23 goes to office-lead review. Every reply requires approval, and none confirms a date.

Prerequisites. Checked draft and rule table.

Proposed time: 35 minutes, including the explanation, practice and review.

TRY A REPRESENTATIVE EXERCISE

Draft a response without making a booking

Fictional practice material

Fictional Meadow Repairs request R-21: “I am in the North area. Can you repair my door on Friday?” Supported areas: North and Central. Selected rule route: prepare draft. Appointment availability has not been checked. Office lead must approve every outbound reply; a staff member sends approved replies.

Example prompt

Draft a brief reply for request R-21 using only the supplied information and selected route. Acknowledge the requested day but do not confirm a booking, price or availability. Include the request ID and label the result “Draft — office-lead approval required.” Do not send anything.

Illustrative expected output

One acceptable example, not a guaranteed AI response.

R-21 — Draft — office-lead approval required. “Thanks for your door repair request in the North area. We have noted your preference for Friday. Appointment availability has not been confirmed yet.”

Check the result. North is supported, but Friday remains a preference. Confirm no booking or price is invented, and the approval label is present. A person decides whether the draft may be sent.

BEFORE WE MEET

What your team provides.

  • A process owner and a description of the current steps.
  • Written rules, exceptions and approval responsibilities.
  • Approved sample requests or permission to use fictional cases throughout.

Use approved, anonymized examples or fictional material. Do not share confidential records or credentials in a planning inquiry.

ADAPT IT TO YOUR WORK

Choose a useful focus.

  • Use an internal purchase request, service inquiry or document intake process.
  • Keep all testing manual for teams still clarifying their process.
  • For teams with an established process, use the test records to scope a later technical review of permissions, integrations and recovery requirements.

We confirm the audience, business task, time and tool access during planning.