A prompt is an assignment. Vague assignments get vague, and sometimes invented, work back. This instructor-led course teaches a repeatable method for directing AI the way a good supervisor directs a new team member. Trainees learn a five-part prompt (role, context, task, constraints, output), how to bind the AI to a source they supply, a four-pass revision routine, and a written accept, revise, or reject decision before any AI text is reused. Every session is hands-on. Trainees rebuild weak prompts, write follow-up passes, strip unsafe customer and company data out of a coworker's prompt, and build a small prompt library for their own recurring tasks. Labs run in the employer's approved tool, such as Microsoft 365 Copilot, ChatGPT, Gemini, or Claude, or on paper if no tool is approved. Examples come from fictional Massachusetts businesses in healthcare, insurance, manufacturing, construction, and hospitality. Trainees leave with a tested prompt library, a one-page prompt card, and a quality-gate habit that protects both their customers and their own name. Suited to any worker who writes, answers questions, or prepares documents and already uses AI at a basic level.
What you will be able to do
By the end of the course, trainees will be able to:
- Build a prompt with all five parts (role, context, task, constraints, output) for a real task type.
- Bind the AI to a supplied source and require 'not in source' for missing facts.
- Revise in passes (structure, content, audience, polish) rather than accepting the first draft.
- Reject an output that does the wrong job, invents a material fact, or sits in a thread containing restricted data, and state why in one sentence.
- Assemble a library of at least five tested prompts for their own recurring tasks.
Who this course is for
Customer service representatives; Administrative staff; Sales and account representatives; HR coordinators; Marketing and communications staff; Office and practice managers
Before you start
Basic experience using any AI chat tool, or completion of FAST-01. No coding.
Hands-on activities
- Lab 1 Rebuild a thin prompt (fictional Cape Harbor Inn volunteer beach cleanup announcement).
- Lab 2 Two passes on a customer reply (rain policy and child-supervision questions answered only from the fact sheet).
- Lab 3 Strip the overshare: remove a customer's address, a child's name and age, and a remark about a coworker; choose Reject.
- Lab 4 Prompt library build and partner swap test.
- Prompt clinic: trainees bring a sanitized prompt from work; the cohort critiques it with the five-part card.
How your learning is measured
- Attendance at every session (opening, start of first lab, dismissal).
- Four lab worksheets scored complete / incomplete against written criteria (all five parts present, source bound, no outside facts, correct gate decision).
- Final: prompt library of 5+ prompts; partner runs two of them and confirms usable output.
