← Part 04: AI tools and prompting, safely PART 04 · AI TOOLS · 25 MINUTES

AI tools and prompting, safely

Present slides Full screen, one slide at a time · 5 slides
MASTERCLASS · PART 04 · AI TOOLS 01

Meet your AI product team

  1. ANALYSTSharpens your problem, customer and assumptions, and asks what you have missed
  2. UX ASSISTANTSuggests user journeys, screens, wording and accessibility fixes
  3. ARCHITECTExplains what a real version needs: data, logins, hosting, security
  4. DEVELOPERBuilds a working prototype screen you can click in the browser
  5. REVIEWERChecks the work against your goals and finds problems, if you ask it to

ChatGPT, Claude and Gemini can play every role. You are the product manager.

MASTERCLASS · PART 04 · SAFE USE 02

Use AI like a smart stranger

Never paste

  • Personal data: names, emails, phone numbers, health or money details
  • Client or customer information, even “just to test”
  • Anything company-confidential, or against your employer’s AI policy

Always do

  • Use invented or anonymised examples, like TutorDesk
  • Check facts, figures and legal claims before you rely on them
  • Make the decisions yourself; the AI suggests, you own it

If you would not read it aloud to a stranger on a train, do not paste it into an AI.

MASTERCLASS · PART 04 · PROMPTING 03

“Build me an app” is not a prompt

Bad prompt

  • “Build me a booking app for tutors.”
  • No users, no problem, no limits
  • The AI guesses everything, confidently

Good prompt

  • CONTEXT: I am planning TutorDesk, a booking tool for independent tutors
  • GOAL and USERS: let parents reschedule a lesson without messaging the tutor
  • CONSTRAINTS and EXPECTED OUTPUT: one screen, phone first, no payments; a one-page brief with open questions

Context, goal, users, constraints, expected output. Five parts, every time.

MASTERCLASS · PART 04 · PROMPTING 04

Build in small steps, not one giant prompt

  1. PROBLEMAsk the AI to challenge your problem statement
  2. VISIONTighten the value proposition and the outcome
  3. FEATURESAgree the smallest set that delivers the value
  4. ARCHITECTUREAsk what a real version would need
  5. PROTOTYPEBuild one screen for one journey
  6. REVIEWAsk it to check the result against your brief

Agree each step, save it in your prompt log, then move on.