FTSM Technopreneurship Bootcamp 2026 · Day 2 · Build

AI-Powered Prototyping

Supercharge your MVP

Ts. Dr. Mohd Nor Akmal Khalid · CAIT, FTSM, UKM

The next 120 minutes

Think first. Then build fast.

10.30 – 11.12

THINK

42 min

Live demo · 4 lenses · Thinking Canvas · Prompt

11.12 – 12.10

BUILD

58 min

Sprint 1 (38) · Peer test (6) · Sprint 2 (14)

12.10 – 12.30

SHOW

20 min

Demos · Reflect

Your goal by 12.30: a tested hypothesis, a live link, a 20-second demo.

● Live demo

One sentence. One working app.

You pick a campus problem
I type one prompt
We keep talking while it builds

> prompt

Build a mobile-first web app where FTSM students book discussion rooms. Show today’s free slots, let a student reserve one with their name, and list “My bookings”. Use realistic sample data.

aistudio.google.com → Build

The map · Double Diamond

You are here: the third quarter

Double Diamond: Design Council (UK), 2004, designcouncil.org.uk/framework-for-innovation

Why MVP, not product

Three stages, three different questions

Prototype

Will they understand it?

Sketches, mock-ups, clickable screens

02 · today

MVP

Will they use it, or pay for it?

The smallest thing a real user can try

Product

Can it scale and earn?

Reliable, secure, supported

Building is no longer the bottleneck.

Knowing what to build — and for whom — is.

What “viable” means

A thin slice, not a wide base

After S. P. Anderson’s UX hierarchy of needs (2011), building on A. Walter (2011); MVP-slice view as drawn by B. Waldock, beingagile.co.uk

How an MVP is born

Design thinking + Lean UX + Agile

Design thinking · understand the need

Lean UX · state and test a hypothesis

Agile · short build–measure–learn loops

Redrawn after a common “Design thinking + Lean UX + Agile = successful MVP” figure. Build–Measure–Learn: E. Ries, The Lean Startup (2011).

Lens 1 · systems thinking

See the system, not just the problem

Lens 2 · design thinking

Start from a person, not a feature

Empathise

Define

Ideate

Prototype

Test

Point of view (POV)

[User] needs a way to [need] because [insight].

e.g. Off-campus final-year students need a way to secure a room before they travel in, because walking in at 2 pm usually fails.

How might we… ? (HMW)

How might we help [user] [do the need] without [today’s pain]?

e.g. How might we help them book a room from home without racing to the faculty?

Five modes after the Stanford d.school; POV and “How might we” are standard design-thinking tools. Dark pills = what today covers.

Lens 3 · computational thinking

Think like a computer before you prompt one

Computational thinking: J. M. Wing (2006), Communications of the ACM 49(3); four-part form as commonly taught.

Lens 4 · Lean UX hypothesis

Make your guess testable

Hypothesis format adapted from J. Gothelf & J. Seiden, Lean UX (2013).

Classic MVPs · real examples

Famous MVPs had almost no code

Explainer video · Dropbox, 2008

Show it before it exists

A 4-minute demo video, posted online. The beta waitlist grew from about 5,000 to about 75,000 overnight.

Wizard of Oz · Zappos, 1999

Humans behind the curtain

Photos of shoes from local stores, posted on a simple site. Each order was bought at the store and shipped by hand.

Fake door · Buffer, 2010

Sell it before you build it

A two-page site, then a pricing page added, before the product existed. Clicks on paid plans showed people would pay.

Sources: YourStory (Feb 2026), Dropbox demo video; Hostinger, “16 best minimum viable product examples”. Illustrations are schematic, not the original products.

The AI shift

What AI changed, and what it didn’t

Changed

A clickable MVP in about an hour
Non-coders can build
Slides-only pitches look weak

Did not change

You still need a real user
You still choose the problem
Judges still ask: who pays?

One laptop per group

Open MVP Studio

Join with your group number

Fill the Thinking Canvas, hints on request

Copy your prompt into AI Studio

12.08: post your share code for the gallery

Scan to open

Worksheet 1 · 12 min · one A3 per group

The Thinking Canvas

1 · SYSTEM

3 min

What keeps the problem in place?

“It keeps happening because…” · “Who else is involved?”

2 · USER

3 min

Point of view + How might we?

“[User] needs a way to … because …” · “How might we …?”

3 · HYPOTHESIS

2 min

Your riskiest assumption, testable

“We believe … will … because …. We’ll know when ….”

4 · SLICE

4 min

Decompose, then cut to 3

List the steps. Circle the 3 that test box 3. Add one “if…then” rule.

Your canvas writes your prompt

From canvas to prompt

CANVAS BOX

PROMPT PART

2 · User + point of view

ROLE & GOAL

4 · Steps of the core job

CORE FLOW

4 · The “if…then” rule

CORE FLOW · rule

4 · The 3 circled steps

FEATURES (only 3)

1 · Who is in the system

LOOK & DATA

3 · Hypothesis signal

PEER TEST at 11.50

Prompt card A · from brief to build

The 4-part build prompt

1 · ROLE & GOAL

Build a web app MVP called [NAME] for [USER] who struggles with [PROBLEM].

2 · CORE FLOW

Opens it → [STEP 1] → [STEP 2] → sees [OUTCOME]. If [X], then [Y].

3 · FEATURES

Only these 3: [FEATURE 1] · [FEATURE 2] · [FEATURE 3].

4 · LOOK & DATA

Mobile-first, [COLOURS]. 15 realistic Malaysian sample records (RM, kolej). Language: [EN / BM].

The build loop

Build, measure, learn, in minutes

One change
per prompt.

first build · 2–5 min
each tweak · 1–3 min
≈ 7 loops in 38 min

Prompt card B · iterate

Five verbs for every next prompt

FIX

When I click [X], [Y] happens. It should [Z]. Fix only this.

CHANGE

Simplify [SCREEN]: keep [A] and [B], remove [C].

ADD

Add one feature: [FEATURE]. Change nothing else.

WOW

Use Gemini inside the app so it [AI FEATURE].

POLISH

Add a landing screen with the tagline “[TAGLINE]”.

Team set-up

Four roles, zero idle time

Builder

Types the prompts

while it generates

Writes the next prompt

User

Clicks and breaks things

while it generates

Writes test steps for the next build

Scribe

Screenshots and pitch notes

while it generates

Drafts the 20-second demo script

Researcher

Guards the brief

while it generates

Prepares sample data and app wording

Rotate the Builder at 11.32. Nobody watches the progress bar.

Sprint 1 · time budget

Your 38 minutes, checkpoint by checkpoint

11.12

Open tool, paste prompt

11.15

v1 prompt submitted

11.20

First preview (2–5 min wait)

11.32

3 loops done · rotate Builder

11.45

Feature freeze: fixes only

11.50

Live link + 3 screenshots

Stuck on one error for two loops? Undo to the last working version and try a smaller prompt.

Your toolkit

One main tool, two backups

MAIN

Google AI Studio · Build

Free for prototyping. Google sign-in. Adds a database and login when asked.

aistudio.google.com

BACKUP

Lovable

Polished look. Free: 5 credits a day; projects are public.

lovable.dev

BACKUP

Bolt.new

Fast, runs in the browser. Free: daily token cap.

bolt.new

Out of credits? Switch account, not tool. Free limits as of Jul 2026; they change often.

Build sprint 1 · 11.12 – 11.50

Build v1

11.15 v1 prompt submitted
11.20 first preview on screen
11.32 3 loops done · rotate Builder
11.45 feature freeze: fixes only
11.50 live link + 3 screenshots

AI generating? Do your role’s job.

Peer test · 11.50 – 11.56

Stand up & swap

One tester per group moves clockwise
2 minutes, no explanation from the team
Team stays silent; note every hesitation
Did your hypothesis signal happen? (box 3)

Use the Peer-test sheet in your kit.

Build sprint 2 · 11.56 – 12.10

Fix + one WOW

First fix where the tester got stuck
Then one WOW, only if v1 works
11.59 WOW prompt submitted
12.06 freeze: nothing new
12.08 post link + screenshots

WOW prompts take longest. Start by 11.59.

Speed demos · 12.10 – 12.22

The 60-second demo

10 s

Problem

Who hurts, and how

30 s

Click through it, live

Show the one core job, start to finish. No feature tour.

10 s

Real vs. mock

Say it plainly

10 s

Tester said…

One line of evidence

4–5 groups demo live. Everyone else: link + screenshots on the shared sheet.

Guardrails

How not to get caught out by judges

No real personal data

No IC, phone or matric numbers. Sample data only (PDPA 2010).

Free projects can be public

Never paste API keys or private UKM data into a prompt.

The app is not the startup

An hour-long build is easy to copy. Your edge: user insight, data, channel.

Keep your prompts

Tools change fast. Your brief and prompts rebuild the app anywhere.

Say what is real.

your pitch line

“Booking works end to end. Payments are mocked. 2 of 3 testers booked without help.”

Reflect · 2 min, in your group

Think about your thinking

SYSTEM

What did the tester show us about the system?

USER

Did they behave the way we believed?

HYPOTHESIS

Confirmed, rejected, or still unclear?

NEXT · DOUBLE DIAMOND

Diverge again, or converge and build more?

Before tomorrow’s pitch

Three things to do tonight

Show it to one real user

Someone outside your group. Watch, don’t explain.

Write a 20-second demo script

Problem, one click-through, what is real.

Bring the link to the pitch

Plus a screen recording, in case the Wi-Fi drops.

Backup slide · no internet

Plan B: build on paper first

Recorded demo

Play the pre-recorded build instead of building live.

Paper prototype

Sketch 4 phone screens on A4. A tester taps; a teammate swaps the paper.

Build tonight

Prompt the AI version at 9.00–10.30 pm. Mentor check at 9.15 am tomorrow.

Build to learn,

not to impress.

akmal@ukm.edu.my · akmal.app · CAIT, FTSM, UKM

AI-Powered Prototyping

Supercharge your MVP

Think first. Then build fast.

THINK

BUILD

SHOW

One sentence. One working app.

You are here: the third quarter

Three stages, three different questions

Prototype

MVP

Product

Building is no longer the bottleneck.

A thin slice, not a wide base

Design thinking + Lean UX + Agile

See the system, not just the problem

Start from a person, not a feature

Think like a computer before you prompt one

Make your guess testable

Famous MVPs had almost no code

Show it before it exists

Humans behind the curtain

Sell it before you build it

What AI changed, and what it didn’t

Changed

Did not change

Open MVP Studio

The Thinking Canvas

What keeps the problem in place?

Point of view + How might we?

Your riskiest assumption, testable

Decompose, then cut to 3

From canvas to prompt

The 4-part build prompt

Build, measure, learn, in minutes

Five verbs for every next prompt

Four roles, zero idle time

Builder

User

Scribe

Researcher

Your 38 minutes, checkpoint by checkpoint

One main tool, two backups

Google AI Studio · Build

Lovable

Bolt.new

Build v1

Stand up & swap

Fix + one WOW

The 60-second demo

Problem

Click through it, live

Real vs. mock

Tester said…

How not to get caught out by judges

No real personal data

Free projects can be public

The app is not the startup

Keep your prompts

Say what is real.

Think about your thinking

Three things to do tonight

Show it to one real user

Write a 20-second demo script

Bring the link to the pitch

Plan B: build on paper first

Recorded demo

Paper prototype

Build tonight

Build to learn,

not to impress.

All slides

Presenting