Skip to main content
RADA Richness Co., Ltd. logo
AI WorkshopNEW About us
Services
Courses
Work
Knowledge Hub
Careers ContactChat with us on LINE
Free marketing consult
Starting a business · Startup with AI

Design Thinking: 5 Stages to Find Solutions Customers Really Want, With Prompts for AI at Every Stage

Before you invest in a new product or service, be clear about the problem customers really have. Design Thinking is a way of thinking that starts with users, then runs small, fast experiments before you invest for real. This article explains the 5 stages visually, tells you what data to prepare for AI, and gives you prompts you can copy into ChatGPT, Gemini or Claude right away.

RADA Richness brand character Creative team · RADA Richness21 November 2026 · 10 min read

What Is Design Thinking

Design Thinking is a human-centered approach to solving problems. It starts by understanding users deeply, then moves on to ideas, prototypes and testing with real users. It grew out of the way designers work. David Kelley, who founded the design firm IDEO in 1991, played a key role in bringing it into business. The most popular 5-stage model, Empathize, Define, Ideate, Prototype and Test, is the one taught by the Hasso Plattner Institute of Design at Stanford University (the d.school).

The 5 Stages of Design Thinking at a Glance

The first two stages help you understand the problem; the last three help you find and try solutions. The curves on top show going back when test results say you need to adjust. Tap a stage to see its explanation and prompt.

Click Copy prompt then paste it into ChatGPT, Gemini or Claude and replace the parts in [square brackets] with your business details. Prompts are written in English so the AI follows instructions precisely, but they tell it to answer in Thai, and you can type your data in Thai

Prompt Guideline: Data to Prepare So AI Can Help With Design Thinking

AI can draft interview questions, summarize patterns and brainstorm ideas very quickly, but it needs data from real users first. Prepare these 8 items and always tell it which stage you are in.

Data to prepareExamples of what to includeWhere to find itFeeds
Users you want to understandWho has the problem, and when, where and with whom they use the productWhere to find itSales history, LINE OA chats, page followersEmpathize
Notes from talking with usersReal quotes from talking with 5 to 8 people, what they do and how they feel (remove names and phone numbers)Where to find itInterviews, phone calls, meetingsEmpathize
What you observedSteps where customers get stuck, ask again, or give up halfwayWhere to find itChats where customers went silent, comments, frequently asked questionsEmpathize
The problem you think it isWhat you currently believe the problem is, written down so AI can challenge itWhere to find itYour own and your team's viewsDefine
Business constraintsBudget, time, people and things you definitely cannot do, so the ideas are realisticWhere to find itYou and your teamIdeate
What you have already triedPast ideas and their results, and how competitors solve the same problemWhere to find itThe shop's experience, competitor pagesIdeate
A prototype you can make this weekSomething quick and cheap, such as a new menu image, chat reply message or landing pageWhere to find itTools you already have, such as Canva or LINE OAPrototype
Test resultsWho tried it, how many, what they did and said, and numbers before and after (with dates)Where to find itChats, short surveys, LINE OA reportsTest

6 rules before you paste data into AI

  1. 01Include real user quotesAI finds patterns well when it gets real words. If you only give it your opinions, the result will just confirm what you already believe.
  2. 02Separate what you saw from what you interpretWrite what customers did or said separately from your opinions, and ask AI to do the same.
  3. 03Say which stage you are inTell AI which of the five stages you are working on, so the answer fits that stage and does not jump to solutions too early.
  4. 04Let AI ask you backTell AI to ask one question at a time when data is missing, rather than guess and give a long answer right away.
  5. 05Remove personal dataRemove customer names, phone numbers and addresses before pasting interview notes into AI.
  6. 06AI is not your real userAI can help draft questions and summarize results, but never use AI's answers in place of talking to and testing with real customers.

Not sure where to start? Let AI interview you first

AI asks one question at a time, tells you which stage you are really in, and what you need to learn from users first.

AI PROMPT
Act as a Design Thinking coach using the 5-stage model (Empathize, Define, Ideate, Prototype, Test).
Interview me ONE question at a time, in Thai, to understand: the business, the user group I want to help, the problem I think they have, what I have heard or seen from real users, and my limits.
If I lack evidence, tell me how to talk to users cheaply this week and move on. When you have enough, tell me which stage I am really in and what to do next. Do not propose solutions yet.

Have data? Go through all 5 stages

A fill-in template covering every stage, giving you insights, a problem statement, ideas, a prototype plan and a test plan.

AI PROMPT
Act as a Design Thinking coach. Take me through all 5 stages using ONLY the data below.

EMPATHIZE
- User group: [PASTE]
- Interview notes and real quotes: [PASTE]
- Observations: [PASTE]

DEFINE
- The problem I think it is: [PASTE]

IDEATE
- Limits (budget, time, people): [PASTE]
- Ideas already tried and results: [PASTE]

PROTOTYPE / TEST
- Tools I have: [PASTE]
- Test results so far, if any: [PASTE]

Then give: 1) key insights about the user, 2) one point-of-view statement and 3 How-might-we questions, 3) 10 ideas and the top 3, 4) a 1-week prototype plan, 5) a simple test plan with what counts as success.
Answer in Thai, in short bullet points a beginner can act on. For every point, say which data it comes from, mark anything that is a guess, and never invent numbers or customer quotes.

The 5 Stages of Design Thinking With Guiding Questions

Dig into each stage with the prompt on its card. The example is a homemade Thai dessert shop selling online via LINE OA, the same fictional shop used in the Business Model Canvas, SWOT and 3Cs articles.

01Understand usersEmpathize

The first stage is going out to understand the people who actually use the product or service: what they do, what they say and the feelings behind it, setting your own assumptions aside. The more real words you hear, the deeper the problems you uncover.

  • What do users do before and after using our product?
  • Where do they get frustrated, hesitate or waste time?
  • Where do they say one thing but do another?
Example: The shop talked with 6 customers who ordered dessert sets for meetings. Most of the people ordering were admin staff who had to guess how many pieces to order for the headcount, and worried about being blamed if there was not enough.
AI PROMPT
I am in the EMPATHIZE stage of Design Thinking.
My business: [WHAT YOU SELL, TO WHOM, WHERE, HOW LONG OPEN]
Who I want to understand: [USER GROUP]
What I already know or suspect: [PASTE]
Write an interview guide of 8 open questions (no yes/no, no leading questions) about their real situation and feelings, plus what I should observe. After I paste my interview notes later, group them into SAYS, DOES, THINKS, FEELS.
Answer in Thai, in short bullet points a beginner can act on. For every point, say which data it comes from, mark anything that is a guess, and never invent numbers or customer quotes.

02Define the problemDefine

Turn what you learned in the first stage into one clear problem statement, written from the user's point of view rather than the company's goals. A popular method is to write a point-of-view statement, then turn it into a "How might we..." question to open the way for ideas.

  • Who is the user we will help?
  • What are they trying to get done?
  • Why does it matter to them (the insight from your conversations)?
Example: Admin staff who order meeting snacks need to order the right amount for the headcount with confidence, because running short or having lots left over makes them look careless to their boss. The resulting question: how might we help them order the right amount in just a few minutes?
AI PROMPT
I am in the DEFINE stage of Design Thinking.
My business: [WHAT YOU SELL, TO WHOM, WHERE, HOW LONG OPEN]
My interview notes and observations: [PASTE]
Find the 3 strongest patterns, then write 3 point-of-view statements in the form [USER] needs [NEED] because [INSIGHT], and turn the best one into 5 "How might we..." questions. Tell me which statement has the most evidence.
Answer in Thai, in short bullet points a beginner can act on. For every point, say which data it comes from, mark anything that is a guess, and never invent numbers or customer quotes.

03Generate ideasIdeate

This stage is about quantity before quality. Come up with many solutions without judging them, including ideas that seem odd, then pick the ideas that best fit users, are doable and give the best value to prototype.

  • Which solutions need no extra investment?
  • If a competitor or another business faced this problem, how would they solve it?
  • Which idea can be tested fastest?
Example: Ideas included dessert sets by headcount (10, 20, 30 people), a calculator chart in the LINE OA menu, individually boxed portions, and saving past orders so customers can reorder in a single message.
AI PROMPT
I am in the IDEATE stage of Design Thinking.
My business: [WHAT YOU SELL, TO WHOM, WHERE, HOW LONG OPEN]
My problem statement / How might we question: [PASTE]
My limits (budget, time, people): [PASTE]
Give me 15 different ideas, from simple to bold, without judging them. Then score each from 1 to 5 on user value, ease for a small team and speed to test, and pick the top 3 with one reason each.
Answer in Thai, in short bullet points a beginner can act on. For every point, say which data it comes from, mark anything that is a guess, and never invent numbers or customer quotes.

04Build a prototypePrototype

Turn the idea into something people can touch, as fast and cheaply as possible. It does not need to be pretty or complete; the goal is for users to try it and give feedback. A prototype can be a sketch, a menu image, a chat reply or a service done by hand for now.

  • What is the smallest piece you need to build for people to try?
  • How many days will it take? Can you use tools you already have?
  • What do you want to learn by having people try this prototype?
Example: Make menu images for dessert sets by headcount in 3 sizes in Canva, plus an auto-reply message in LINE OA. It takes half a day, with no changes to the real ordering system yet.
AI PROMPT
I am in the PROTOTYPE stage of Design Thinking.
My business: [WHAT YOU SELL, TO WHOM, WHERE, HOW LONG OPEN]
The idea I chose: [PASTE]
Tools I already have: [E.G. CANVA, LINE OA, GOOGLE FORMS]
Suggest the cheapest prototype I can make in 1 to 3 days that real users can try. List what to build, what to fake or do by hand for now, and the 2 or 3 things I want to learn from it.
Answer in Thai, in short bullet points a beginner can act on. For every point, say which data it comes from, mark anything that is a guess, and never invent numbers or customer quotes.

05TestTest

Let real users try the prototype, watch what they do and say, then take the results back to improve. Test results often send you back to an earlier stage, sometimes all the way to redefining the problem, which is a normal part of the process.

  • Can users do it on their own without explanation?
  • Where do they get stuck, and what do they say while trying it?
  • Does this result tell us to adjust the prototype, the idea or the problem?
Example: The shop sent the new menu to 10 past customers. Most ordered faster, but many asked about receipts in the company's name, so the shop went back to Define and found that claiming expenses was another problem customers had.
AI PROMPT
I am in the TEST stage of Design Thinking.
My business: [WHAT YOU SELL, TO WHOM, WHERE, HOW LONG OPEN]
My prototype: [DESCRIBE]
Who tried it and what happened (actions, quotes, numbers with dates): [PASTE]
Summarise what worked, what failed and what surprised users. Then tell me which stage to go back to (Empathize, Define, Ideate or Prototype) and exactly what to change next round.
Answer in Thai, in short bullet points a beginner can act on. For every point, say which data it comes from, mark anything that is a guess, and never invent numbers or customer quotes.

Design Thinking Is Not a Straight Line: Iterate Until You Find What Works

The five stages are a guide, not a staircase you must climb one step at a time. Teams can work on several stages at once, and test results often send you back to an earlier stage. See how the dessert shop went through two rounds.

Round 1 · EmpathizeTalked with 6 customers who ordered dessert sets for meetings and kept hearing that they feared ordering the wrong amount for the headcount
DefineProblem: how might we help people ordering meeting snacks order the right amount in just a few minutes?
IdeateCame up with 15 ideas and chose dessert sets by headcount, because it is fast and needs no extra investment
PrototypeMade menu images in 3 sizes in Canva and an auto-reply message in LINE OA in half a day
Test10 past customers tried ordering; they ordered faster, but many asked about receipts in the company's name
Round 2 · Back to DefineThe new problem statement includes customers' expense claims, then a new round of ideas and testing begins

Plan a user test

Get a 5-step test script, neutral questions, a results table and criteria for which stage to go back to.

AI PROMPT
Help me plan a small user test for my prototype.
My business: [WHAT YOU SELL, TO WHOM, WHERE, HOW LONG OPEN]
Prototype: [DESCRIBE]  What I want to learn: [PASTE]
Users I can reach: [WHO AND HOW MANY]
Write: who to invite and how, a 5-step test script with neutral questions, what to observe, a simple results table, and what result would make me go back to Define, Ideate or Prototype.
Answer in Thai, in short bullet points a beginner can act on. For every point, say which data it comes from, mark anything that is a guess, and never invent numbers or customer quotes.

Let AI check your work

Find places where you used beliefs instead of real data, or jumped to solutions too early.

AI PROMPT
Here is my Design Thinking work so far:
[PASTE]
Check it strictly: did I talk to real users or rely on my assumptions? Is my problem written from the user's point of view? Did I jump to a solution too early? Is my prototype cheap enough to throw away? Point out weak spots and what to do before the next round.
Answer in Thai, in short bullet points a beginner can act on. For every point, say which data it comes from, mark anything that is a guess, and never invent numbers or customer quotes.

What you learn in the Empathize and Define stages can go straight into the customer side of the Value Proposition Canvas Ideas that pass testing can then be built into a business model in the Business Model Canvas and compared with competitors using the 3Cs Model

Mistakes beginners often make

  • Skipping the Empathize stage and starting from an idea you like yourself
  • Asking customers "Do you like it?" or "Would you buy it?" instead of asking about what they actually do
  • Writing the problem as the shop's goal, such as wanting to sell more, instead of the user's problem
  • Making the prototype too polished and expensive, so you cannot bear to throw it away when test results are poor
  • Thinking you go through the five stages once and you are done, when you need several rounds of iteration
  • Using AI's answers in place of real customers' voices

Checklist before you use it

  • Talked to or observed at least 5 real users before defining the problem
  • The problem is written from the user's point of view and backed by evidence from conversations
  • Had several ideas before choosing, not just the first idea that came up
  • The prototype is quick and cheap, and you are ready to throw it away if results are poor
  • Tested with real users, noted what they did and said, then decided which stage to go back to
  • Removed personal data before pasting into AI, with no numbers or customer quotes invented by AI

References: Interaction Design Foundation: 5 Stages in the Design Thinking Process, Stanford d.school: Design Thinking Bootleg, History and concepts of Design Thinking, IDEO U: What is Design ThinkingThere are many models, from 3 to 7 stages; this article uses the d.school's 5-stage model. The example business in this article is fictional. AI output is a suggestion to consider; test with real customers before making investment decisions.

Was this guide useful?
Share it with a friend who is starting a business.
Share to Facebook Share on LINE