BADM 346 · Nov 19, 2026 · Guest workshop

The Moat Sprint: your product is not your moat.

Nine AI agents built a prototype for each team in about five minutes. Now find what an AI cannot copy.

How this works with any AI tool

Use whatever you have: ChatGPT, Claude, Gemini, Copilot, NotebookLM or anything else. No installs, works on a phone.

Same input, same prompt, different tools. Each teammate pastes the same prompt and the same brief into a different tool. Each prompt asks for a fixed format, so you can put the answers side by side.

Where the tools agree, it is probably solid. Where they disagree, that is what your team should argue about.

Everything is text. If a tool cannot open links or files, paste the brief instead. No AI at all? Answer the same questions on paper.

The four steps (35 minutes)

  1. 5 min · Open your prototype and break it. One person plays the customer. One plays the skeptic. Note one thing the AI got right that you had not thought of. Find your team.
  2. 10 min · Prompt A, the copycat test. What could a competitor copy in a weekend? What not in six months?
  3. 10 min · Prompt B, score your moats. Six kinds of advantage, scored 0 to 3. Pick one to build first.
  4. 10 min · Prompt C, a 7-day test. One cheap, no-code test that could prove your moat wrong.

The prompts

Step 2 · 10 min

Prompt A: The copycat test

Act as a well-funded competitor who saw our venture today.

[PASTE YOUR TEAM'S BRIEF HERE]

1. In a table, list what you could copy in one weekend with AI tools, and what you could NOT copy in six months, and why.
2. Name the single hardest thing for you to copy.
Be specific to our venture. Do not flatter us.

Step 3 · 10 min

Prompt B: Score your moats

Here are six kinds of advantage: network, trust, distribution, data, operations, switching cost.

For our venture (brief below), score each from 0 to 3 for how strong it could become in our first year. For each one, give one sentence of evidence and one risk.
Then tell us which ONE we should build first, and why. Be skeptical.

[PASTE YOUR TEAM'S BRIEF HERE]

Step 4 · 10 min

Prompt C: A 7-day test

We think our moat is [YOUR PICK]. Design one test we can run in the next 7 days, on campus, for under $50, with no code, that would show whether this moat is real.
Give: what we do, who we talk to, what number we count, and the result that would prove us wrong.

Our venture: [PASTE YOUR TEAM'S BRIEF HERE]

Your moat card

Anyone can build our app.
What they cannot easily copy is  ______ ,
because  ______ .
We will know it is real by [date] if [number] happens.

Your team's prototype and brief

These prototypes were built by AI agents (OpenAI Codex) from one-page briefs of your September decks. All data in them is invented. They are yours: open one, save the page, and keep building in any tool.

FlashCab

Open the FlashCab prototype

What the agent built: A dark, mobile-friendly FlashCab prototype lets riders choose a neighborhood, compare shared cars and seat fares, book a seat, review pickup and stop details, share a trip, rate the ride, and switch to a driver earnings view.

The brief the agent got (paste this into your prompts)

Pitch

College students coming home from a night out pay too much for a taxi or wait too long. FlashCab sells single seats in a shared car that has 3 to 6 seats, so the fare is split between riders going the same way. Bars near campus give discounts to students who ride FlashCab.

User

A college student leaving a campus bar late at night who wants a cheap, safe ride home.

Core flow to build

  1. The rider enters a drop-off area (choose from 6 to 8 invented campus neighborhoods).
  2. The app shows cars heading toward that area. Each car shows its seats (filled and empty), the ETA and the current price per seat.
  3. The rider taps an empty seat to book it. The price per seat drops as more seats fill, and the app shows how much the rider saves compared with a solo ride.
  4. A confirmation screen shows the driver's first name, the car, the driver rating and the stop order (where the rider is in the drop-off sequence).
  5. Safety controls are visible: a "women-only car" filter, share-trip, and rating after the ride.

Nice to have

  • A toggle to the driver view that shows earnings for the full car compared with one solo fare.
  • A partner-bar discount banner.

Tone

Fast, night-time, high-contrast. A dark theme with one bright accent color.

GlobeU

Open the GlobeU prototype

What the agent built: Clickable offline journey from mock school verification through profile setup to student matches, city guide, and Lisbon group chat.

The brief the agent got (paste this into your prompts)

Pitch

Students going abroad arrive in a new city where they know no one. GlobeU connects students from different universities who will study in the same city in the same semester, and it gives practical help on visas, insurance, housing and daily life.

User

A college student aged 18 to 24 going abroad for one semester who does not know anyone else going to the same city.

Core flow to build

  1. Sign-in screen with "Choose your school" (mock school-account verification; show a verified badge after sign-in).
  2. A short profile: destination city, semester, home university, interests (chips), lifestyle (early bird or night owl, quiet or social) and whether they need a roommate.
  3. A matches screen: students going to the same city and semester, each with a compatibility score and the reasons for it ("both night owls, both like hiking"). There are filters for Friends, Roommates and Travel partners.
  4. A city guide tab for the chosen city: a checklist for visa, insurance, housing and transport, with items to tick off.
  5. A group chat preview for "Everyone in <city>, Spring 2027".

Nice to have

  • A countdown to the departure date.
  • A suggested first-week meetup event.

Tone

Warm, friendly and reassuring. A travel feel with soft colors.

DesignBridge

Open the DesignBridge prototype

What the agent built: A responsive DesignBridge prototype takes renters from a room brief through five student concepts to a budgeted favorite sheet with a review form.

The brief the agent got (paste this into your prompts)

Pitch

Young renters have Pinterest taste but cannot pay a professional interior designer. Interior design and architecture students need real clients for their portfolios. DesignBridge is an open forum where students design rooms for real briefs, and clients pick the concept that fits their taste and budget.

Users

  • Demand side: a person aged 20 to 35 furnishing a dorm, a first apartment or a first home on a budget.
  • Supply side: a design or architecture student who is building a portfolio.

Core flow to build (client side)

  1. Post a room brief: room type, dimensions, budget (slider), style (chips such as Japandi, Mid-century, Boho or Industrial) and a short note.
  2. Concepts arrive: 4 to 6 concept cards from invented student designers. Each card has a CSS or SVG mood board (color palette swatches and simple furniture outlines), the designer's school and year, the style tags and an estimated total cost.
  3. Filter or sort the concepts by budget fit and style.
  4. Choose a favorite, then see the full concept sheet: the palette, a furniture list with prices that adds up to a total against the budget, implementation tips and a "Leave a review for the designer" box.

Nice to have

  • A toggle to the designer view: open briefs, and a portfolio counter ("3 real client projects").

Tone

Clean, editorial and design-forward. Plenty of white space.

Lockstep

Open the Lockstep prototype

What the agent built: A responsive Lockstep student prototype that checks mock Canvas tasks, adds personal tasks, unlocks focus apps after three verifications, and starts an earned-time timer.

The brief the agent got (paste this into your prompts)

Pitch

Students get distracted on their devices and forget assignments. Website blockers exist, but they are not tied to the work the student actually has to do. Lockstep locks distracting apps and sites until the student's task list is done, and it checks completion against the learning platform (Canvas or Google Classroom) instead of trusting a tick box.

User

A high school or college student on a school-issued device. The buyer is the school district.

Core flow to build (student view)

  1. A "Today" screen with tasks synced from a mock learning platform (for example "Chem lab report, due 11:59 PM"), plus tasks the student adds.
  2. A grid of locked apps (invented apps with emoji icons, such as a video app, a game or a social feed). Each shows a lock icon and "Unlocks when 3 of 4 tasks are done."
  3. The student marks a task done. The app shows "Checking with Canvas…" and then verified or "not submitted yet" (mix both outcomes in the mock data).
  4. A progress ring fills up. When the goal is met, the apps unlock with a clear, satisfying animation and a timer for earned free time.

Nice to have

  • A toggle to a teacher or district view: the class completion rate and the minutes earned.
  • A streak counter.

Tone

Focused and motivating, like a game reward, but calm and not childish.

WhereNext

Open the WhereNext prototype

What the agent built: Offline semester travel planner with schedule-aware weekends, budget-fit trips, itineraries, group joining, and trip spending.

The brief the agent got (paste this into your prompts)

Pitch

Study-abroad students in Europe want to travel every free weekend. Planning is spread over many booking sites, prices keep changing, and it is hard to find others to travel with. WhereNext plans the whole semester of weekend trips around the student's class schedule and budget, and it connects them with other students going to the same place.

The founders claim that a general AI chat cannot do these things: plan around the academic schedule, compare travel options, keep the bookings in one place, or connect students. Build exactly those things.

User

A US college student studying abroad in Europe for one semester.

Core flow to build

  1. Enter details: host city (choose from 6 European cities), semester dates, the weekdays with classes (checkboxes), a budget per trip and travel style (chips).
  2. A semester calendar that shows every free weekend. Weekends with classes or exams are greyed out.
  3. For a chosen weekend, show 3 trip options (invented destinations with mock train or flight times, prices and hostel prices) that fit the budget.
  4. Pick one. The app builds a day-by-day itinerary and shows "4 other students from your city are going" with a Join button.
  5. A "My semester" view with all planned trips and the total spent against the semester budget.

Nice to have

  • A sorting toggle for "cheapest" and "most popular with students".

Tone

Adventurous and bright, with a map and travel feel. Make it confident, because this team is pitching against ChatGPT.

MoveMate

Open the MoveMate prototype

What the agent built: An interactive first-apartment planner with a live budget estimate, move-in cash plan, 30-day checklist, priced essentials list, and roommate bill settlements.

The brief the agent got (paste this into your prompts)

Pitch

A first apartment is confusing. Apartment apps help you find a place, then leave you alone with the deposit, the lease, utilities, furniture and bills. MoveMate guides first-time renters through the whole move: budget, checklist, timeline and splitting bills with roommates.

User

A college student, recent graduate or young adult renting a first apartment. This is the "what am I forgetting?" group.

Core flow to build

  1. Budget check: monthly income, rent and number of roommates. The app estimates utilities, internet, groceries and renter's insurance, shows the money left over, and gives a clear "affordable / tight / too much" signal. It also shows the move-in cash needed (deposit, first month and setup costs).
  2. 30-day move-in timeline: a checklist grouped by week (sign the lease, book movers, set up electricity and internet, buy essentials). Items tick off and a progress bar fills.
  3. First-apartment shopping list with the cost of each item and a total.
  4. Roommate bill split: add bills, choose an even or custom split, and see who owes whom.

Nice to have

  • Lease basics: 5 terms explained in plain English.
  • Reminders for upcoming due dates.

Tone

Calm, organized and friendly. It should feel like a helpful older sibling.

Evergo

Open the Evergo prototype

What the agent built: Click-through family booking, accessible group seat map, ride-day check-in timeline and large-text senior view for Evergo.

The brief the agent got (paste this into your prompts)

Pitch

Many seniors stop driving and feel unsafe on transit or with rideshare apps. Local exercise, education and social programs exist, but seniors cannot reach them, so they become isolated. Evergo runs wheelchair-accessible group rides (1 wheelchair place plus 3 to 4 seats) to partner programs. Local businesses pay a fee for each senior delivered, and adult children get peace of mind.

Users

  • The senior, who rides.
  • The adult child, who often books and wants updates.
  • The local partner (pharmacy, senior center, fitness class), who pays.

Core flow to build (family view first)

  1. Book for a parent: choose the parent's profile (it shows a wheelchair need and the home address), then choose from a weekly list of partner programs (for example "Chair yoga, Tue 10 AM, Community Center").
  2. Group ride view: the van's seat map (one wheelchair place plus 4 seats), who else from the area is going (first names only) and the pickup time.
  3. Ride-day alerts: a timeline of "Picked up 9:42", "Arrived 9:58" and "Heading home 11:15", shown as a check-in feed for the family.
  4. A very simple senior view with large text and big buttons: "My next ride: Tuesday, Chair yoga" and a Call button.

Nice to have

  • A partner view: seniors delivered this month and the fee owed.

Tone

Trustworthy, warm and highly accessible. Use large type and high contrast, and design the senior view for older eyes.

SkillSwap

Open the SkillSwap prototype

What the agent built: UIUC students verify with a mock ID, post a campus task, choose a skilled helper, track the job, review the fee and payout, and rate the swap.

The brief the agent got (paste this into your prompts)

Pitch

Students need help with small tasks (changing a shower head, fixing a cracked phone screen, assembling furniture) that are too small or too expensive for a professional. Other students have exactly these practical skills. SkillSwap is a marketplace for these tasks, limited to verified students at one campus. SkillSwap takes a 10% commission.

Users

Students who need help, and students with skills who want to earn money. Both are verified with a student ID. The first campus is UIUC.

Core flow to build

  1. Sign up with a student ID (a mock verification and a verified badge).
  2. Post a task: category (Home repair, Tech or digital repair, Moving, Tutoring and more), description, location on campus, time window and the price offered.
  3. Students with that skill are notified. Show 3 to 4 helper cards with their skill certificate, rating, completed tasks and price.
  4. A helper accepts. A task tracker shows Accepted, On the way, Done.
  5. Pay: a receipt that shows the task price, the 10% SkillSwap fee and the helper's payout, then a rating.

Nice to have

  • A toggle to the helper view: a feed of open tasks that match my skills and my earnings this month.
  • Safety: in-app chat only, and campus meeting spots.

Tone

Energetic, campus-y and practical. Make it feel peer to peer, not corporate.

TeeUp

Open the TeeUp prototype

What the agent built: An offline TeeUp prototype to log a golf round, calculate its score differential, collect two flight partner confirmations, browse leaderboards, and find golfers.

The brief the agent got (paste this into your prompts)

Pitch

Golfers find it hard to find playing partners at home or while traveling. Golf networks stay local to one club. Existing apps do score tracking and GPS only, with no community. TeeUp combines fair, verified competition (based on handicap, course rating and slope rating) with a social network for golfers.

User

Recreational golfers, ambitious amateurs, young golfers and business networkers. The MVP is tested at two partner clubs in Munich.

Core flow to build

  1. Log a round: choose a course (invented courses with course and slope ratings), enter the gross score and the handicap index. The app computes the score differential with the standard formula, (113 / slope) × (gross − course rating), and shows it.
  2. Flight confirmation: the round shows "Waiting for confirmation from 2 flight partners", which turns into Verified with a badge.
  3. Leaderboards with tabs for My club, Germany and Global, ranked by score differential so players on different courses can be compared. Include a weekly challenge banner.
  4. Find a partner: filter golfers by handicap range, location and a "business networking" tag. Each card shows the handicap, home club and a Request a round button.

Nice to have

  • An activity feed ("Lena shot 79 at a partner club, verified").

Tone

Premium, sporty and clean, with green on white. It should look like an app for golfers with money to spend.