Side HustlesIntermediate23 min read

No-code or code? Choose your MVP type

Four kinds of first version from a manual service to a small coded app, choosing by evidence, hours and risk, where AI helps and adds risk, cutting to one job and three must-haves, a safety and honesty checklist, and a 14-day scope.

No-code or code? Choose your MVP type

No-Code or Code? Choose Your MVP Type

You have talked to customers and run a small test, and the signs are good enough to build something. This is the moment many solo founders go wrong. They open a code editor, or a no-code builder, or an AI app generator, and spend three weeks on a product with login screens, dashboards and settings, for a customer who only wanted one result. The question is not "which tool is best?". It is "what is the smallest first version that can earn, and what is the cheapest honest way to deliver it?". Often the answer involves no software at all. This tutorial shows four kinds of first version, from a manual service to a small app, gives you a way to choose among them by looking at your idea, your hours and your skills, explains where AI genuinely helps and where it creates risk, and ends with a one-page scope and a rule: ship something a real customer can use within 14 days.

Difficulty: ★★☆☆☆ (no coding needed to follow; some comfort with simple tools helps)
Tools needed: A notes app, and any chat AI for the exercises
Last updated: November 2026

Overview

An MVP, short for minimum viable product, is the smallest version of your offer that a real customer can use and pay for, so that you can learn from it. The word "viable" matters: it must actually deliver the promised result. The word "minimum" matters more: every extra feature is a cost, a delay and a place for bugs.

A common confusion is that an MVP must be software. It does not. For many one-person companies, the best first version is a service you perform by hand, a document you send or a simple page with a form. The aim is to find out whether customers want the result. How the result is produced can be improved later, once someone is paying.

AI has made two things easier at once: building small software, and building too much. A chat AI can write a working page or a small tool in an afternoon. That is real. It also means a beginner can generate a lot of code they do not understand, store customers' data insecurely, break something and not know why. A fast build is not a good reason to build. The test is still whether it is the smallest thing that serves the first customer.

This tutorial has seven steps. First, what an MVP is for. Second, the four types of first version. Third, how to choose by looking at your idea, hours and skills. Fourth, where AI helps in each type, and the risks. Fifth, what to cut: the must-haves and the "not now" list. Sixth, a basic safety and honesty checklist, especially if you collect data or take payments. Seventh, writing a one-page scope with a "done means" and a ship date.

A note on honesty. This tutorial names no specific tools and gives no prices, because tools and terms change quickly. Nothing here is legal, tax or security advice. Where a customer's data, money or safety is involved, check current official guidance, the tools' own terms and, if needed, a professional.

Who This Is Useful For

  • People whose idea has early evidence and who are about to build.
  • People tempted to start with a full app, an elaborate site or an automation.
  • Anyone who wants to ship within two weeks, with limited hours.
  • What You Will Learn

  • What an MVP is, and what it is not.
  • Four types of first version: manual service, document or template, no-code page or tool, and a small coded app.
  • How to choose by your idea, your hours and your skills.
  • Where AI helps in each type, and where it adds risk.
  • How to cut to one job and at most three must-haves.
  • A safety and honesty checklist for data, payments and claims.
  • How to write a one-page scope with a ship date.
  • What You Need

    1. Your validation plan or smoke test result, and your limits sheet, if you have them. If not, one sentence on the offer and your honest weekly hours.
    2. A notes app.
    3. Forty-five minutes for the exercises.
    4. A chat AI, optional.

    Information as of 14 November 2026. Tools, features, prices and terms change often. This tutorial names none. Check the current terms and privacy policy of any tool you use before putting customers' information into it.

    Step 1 — What an MVP is for

    An MVP is a learning tool that happens to earn. Its job is to answer questions that interviews and smoke tests could not:

  • Will people actually use it, not just sign up?
  • Will they pay for the real thing, once it exists?
  • Where do they get stuck?
  • What do they ask for next?
  • Does it cost me more time and money than I expected?
  • What an MVP is not.

  • It is not a prototype you show friends. A real customer should use it for a real purpose.
  • It is not a finished product. It will be rough, and that is fine if it is honest.
  • It is not the whole vision. Your vision is a direction. The MVP is the first step.
  • It is not a polished brand. A plain, clear delivery beats a beautiful, late one.
  • The "one job" rule. A good MVP does one job for one kind of person. "Helps freelancers keep a clean monthly income and expense sheet" is one job. "Manages your whole business" is not. Write the job in one sentence. If you need the word "and" twice, it is probably two products.

    The "first customer" test. Imagine your first paying customer. What is the one thing you would want to have in their hands at the end? That is the MVP. Everything else is optional.

    A word on speed. Aim to put something in front of a real customer within 14 days of deciding to build. That is not a magic number. It is a discipline that forces you to cut. If it cannot ship in two weeks of your hours, the scope is too big, or the lane or the idea needs rethinking.

    Why small wins. A small version gives feedback while the stakes are low. A big version delays feedback and raises your sunk cost, which makes you less willing to change.


    Step 2 — The four types of first version

    There are four common types. They run from the lightest to the heaviest.

    Type 1: Manual service ("concierge"). You deliver the result by hand, one customer at a time. There may be no software at all, only a form to collect details and your own work.

  • Examples: you tidy a freelancer's monthly records yourself; you write a personalised weekly meal plan by hand; you review a resume and send comments.
  • Strengths: fastest to start, cheapest, and the most learning. You see exactly what customers need, in detail.
  • Weaknesses: it does not scale. Your hours limit your customers.
  • Good for: services, and any idea where you are not yet sure what the result should be.
  • Type 2: Document or template. You package your knowledge or work as a file: a spreadsheet, a printable pack, a guide, a checklist, a prompt pack.

  • Examples: a bookkeeping spreadsheet; a printable meal-plan pack; a short trail guide as a PDF.
  • Strengths: make once, sell many times, no software to maintain, simple to deliver.
  • Weaknesses: crowded market, and customers may want more help than a file gives.
  • Good for: products, and for turning a service's repeated work into something reusable.
  • Type 3: No-code page or tool. You assemble a working experience from existing building blocks: a page, a form, a booking tool, a spreadsheet with automation, a membership or payment tool, connected with simple rules.

  • Examples: a booking page with a payment step; a form that sends a personalised email; a spreadsheet that calculates a quote; a members-only page.
  • Strengths: looks and works like a real product, with little or no code, and you can change it fast.
  • Weaknesses: you depend on the tools' terms, limits and prices, and complicated logic gets messy. Data you collect has to be handled carefully.
  • Good for: services that need booking and payment, simple tools and early software tests.
  • Type 4: Small coded app. You build a custom application, with or without AI writing the code.

  • Examples: a calculator with user accounts; a tool that processes uploaded files; a custom dashboard.
  • Strengths: full control, and it can do things no ready-made tool can.
  • Weaknesses: the most time, the most places for bugs and security problems, and ongoing maintenance. With AI-written code, you may not understand what you have, which makes problems harder to fix.
  • Good for: a problem proven by earlier steps that really needs custom software, and where you or someone you trust can maintain it.
  • A comparison.

    Manual serviceDocument / templateNo-code page or toolSmall coded app
    Time to first customerDaysDays to a couple of weeksOne to two weeksWeeks
    Upfront costLowestLowLow to mediumMedium
    Learning from customersHighestMediumMediumMedium
    Scales without youNoYesPartlyYes
    Technical riskNoneLowMediumHighest
    Data and security dutiesLow to mediumLowMediumHighest
    Skill you needYour craftYour craft and makingTool skillsCoding or a trusted developer

    Mixed types are common and good. A typical path: start with a manual service (type 1), notice the repeated work, make a template (type 2), then add a booking and payment page (type 3), and only later, if customers ask and pay, consider a small app (type 4). Each step is funded by the one before.

    Step 3 — How to choose

    Choose by looking at four things, in this order.

    1. What does your evidence say the customer wants? Look at your interviews and tests. Do customers want a result (a tidy record, a plan, a decision) or a tool (something they use themselves)? If they want a result, a manual service can deliver it. If they said "I want to do this myself, faster", a tool is closer to what they need.

    2. How well do you understand the process? If you cannot do the job by hand yet, you cannot automate it. Doing it manually for a few customers teaches you the steps, the edge cases and what to leave out. Building software before that is guessing.

    3. What are your hours and skills? Use your limits sheet. With about five good hours a week, a manual service or a template is realistic in two weeks. A coded app probably is not. Be honest about skills. If you have never built an app, your first app is likely to take longer than you think, and AI-generated code you cannot read will be hard to fix.

    4. What are the risks? If the first version will hold other people's data, money or sensitive information, the cost of mistakes goes up. Prefer a simpler type, or established tools designed for that job, over building your own.

    A decision path.

    Can a customer get the result if I do the work by hand?
      Yes → Start with a MANUAL SERVICE (type 1). Ship in days.
      No → Is the result a file or a guide they can use alone?
         Yes → DOCUMENT / TEMPLATE (type 2).
         No → Can existing tools (page, form, booking, payment, spreadsheet) deliver it?
            Yes → NO-CODE (type 3).
            No → Is the problem proven (paying customers or pre-orders) and can I or a trusted
                 person build and maintain it?
                  Yes → Small CODED APP (type 4), scoped to one job.
                  No → Go back a type. Prove it by hand first.

    Signs you are choosing too heavy a type.

  • You are designing features nobody has asked for.
  • You have no paying customer yet, and you are setting up login and user accounts.
  • You cannot describe the first customer's result in one sentence.
  • You are choosing a technology before choosing a job.
  • You expect to finish "after a few more weeks".
  • Signs you are choosing right.

  • A customer could use it by the end of this fortnight.
  • It does one job.
  • You could explain every part of it to someone else.
  • If it failed tomorrow, you could deliver the result by hand.
  • An example. Wen offers a bookkeeping tidy-up. She is tempted to build an app that automatically categorises receipts. After her smoke test she has two interested clients. Her evidence says clients want a result: a clean monthly sheet. Her hours are eight a week. She chooses type 1 with a bit of type 3: she will tidy records by hand, using a shared spreadsheet template, with a simple form to collect receipts and a booking page for a monthly call. She will consider a tool only after she has done the process for five clients and knows exactly which step takes the most time.

    Step 4 — Where AI helps, and where it adds risk

    AI changes what each type costs you, and what can go wrong.

    In a manual service.

  • Helps: drafting your messages and reports, summarising customer details, building checklists, preparing templates, checking your work for gaps.
  • Risk: pasting customers' private details into tools you have not checked. Sending an AI-written message you did not read.
  • Rule: you do the work and take responsibility. AI prepares.
  • In a document or template.

  • Helps: drafting the content, designing a layout, producing variations and formatting.
  • Risk: errors in the content, copied material, and a product that looks like a thousand others.
  • Rule: test it yourself. Use it yourself before you sell it. Check facts and rights.
  • In a no-code page or tool.

  • Helps: suggesting how to connect the tools, writing the page text, writing formulas, drafting email templates and explaining settings.
  • Risk: misconfigured forms that expose data, links that let others see customers' details, forgotten test entries and tools whose terms do not allow your use.
  • Rule: test as if you were a customer, and as if you were a stranger. Check who can see what.
  • In a small coded app.

  • Helps: writing a first version of the code, explaining errors, suggesting tests, writing documentation.
  • Risk: code you cannot read or fix, hidden bugs, insecure handling of passwords or data, dependencies that break and unclear costs when it runs. A working demo is not a safe product.
  • Rule: keep the scope tiny. Do not store sensitive data in your first version. Use established tools for login, payment and data storage instead of writing your own. Test it, read what it does and have someone experienced review it if customers' information is involved.
  • A habit that helps with every type. After AI produces something, ask it to explain it back: "Explain in plain language what each part does, what could go wrong and what data it touches." If you cannot follow the explanation, you are not ready to ship it.

    Use AI to compare the options for your idea.

    Role: a careful advisor for a first-time solo founder. You do not know my customers. Do not invent facts or recommend specific paid products.
    Task: I will describe my offer, my customer's result, my weekly hours and my skills. (1) Explain which of these first-version types fits best and why: manual service, document or template, no-code page or tool, small coded app. (2) Name the single biggest risk in that type for my case. (3) Describe the smallest version I could ship in 14 days, and what I would do by hand. (4) List anything that involves other people's data or money that I should be careful with.
    [paste your offer, result, hours, skills]

    Read the answer critically. You know your customers. The AI does not.

    Step 5 — Cut to one job and three must-haves

    Scope is where most MVPs go wrong. Everything seems important. Cut until it hurts, then cut once more.

    Step A: write the one job.

    THE ONE JOB
    [Specific customer] gets [specific result] by [how], in [time].

    If you cannot write it in one sentence, you are not ready to build.

    Step B: list everything you could build. Write every feature, page, setting and idea. Do not filter yet.

    Step C: apply three filters.

    1. Would the first customer pay without it? If yes, it is not a must-have.
    2. Can I do it by hand instead? If yes, do it by hand for now.
    3. Does it serve the one job? If not, cut it.

    Step D: keep at most three must-haves. For example, for a bookkeeping tidy-up: (1) a way to receive the customer's receipts, (2) the clean monthly sheet, (3) a short call or message to explain it. Everything else waits.

    Step E: write the "not now" list. At least two things you will not build in this version. Typical candidates:

  • Automation and integrations
  • A custom website or branding
  • User accounts and dashboards
  • Extra features and options
  • Mobile apps
  • Perfect design
  • Write them down so they stop nagging you. They are not lost. They are parked until customers ask for them.

    Step F: define "done means" with a number you can check. For example, "three paying clients have received their first monthly sheet", or "20 people have used the calculator and 5 have paid". Done is not "I feel finished". It is a fact.

    Step G: set a ship date within 14 days. Work backwards from your honest weekly hours. If the build needs 20 hours and you have 6 good hours a week, that is about three and a half weeks, and the scope is too big. Cut it.

    The "by hand" principle. Anything that happens behind the scenes can be done manually in version one. The customer does not see it. They see the result. Manual work is cheap to change and teaches you what to automate later.

    Do not hide manual work dishonestly. You do not need to say "a human is doing this" on every line, but do not claim automation or AI that does not exist. If the customer asks, tell them. Being honest about how it works builds trust.

    Write your scope.

    MVP SCOPE
    Date: ___   Idea: ___   Type (1-4): ___
    The one job (one sentence): ___
    Must-haves (max 3): 1. ___  2. ___  3. ___
    Not now (at least 2): ___
    How it is delivered: ___
    What I do by hand: ___
    Done means (a number I can check): ___
    Hours available per week: ___   Days to ship (max 14): ___
    Biggest risk: ___   How I will limit it: ___

    This is the page you will write in this module's quest.

    Step 6 — A safety and honesty checklist

    A small first version still has real customers, so it still carries duties. Check these before launch. They are general principles, not legal or security advice, and rules differ by country.

    Data.

  • Collect the least you need. Every piece of data is a duty. Do not ask for what you do not use.
  • Keep sensitive data out of the first version if you can: identity numbers, bank logins, health information, children's details. If the work needs them, use established tools designed for that, and check the rules where you live.
  • Know where data lives and who can see it. Check the sharing settings of every form, sheet and folder. Open them as a stranger would.
  • Check the terms of every tool before you put customers' information in it, including AI tools. Some tools keep or learn from what you give them.
  • Tell customers what you do with their information, in plain words, and how to ask for deletion.
  • Delete data you no longer need.
  • Payments.

  • Use an established payment tool. Never ask customers to send card details by message, and never store card numbers yourself.
  • State the price, what is included, delivery time and refund terms before payment.
  • Keep records of every sale and refund.
  • Check tax and registration rules for selling where you live.
  • Honesty and claims.

  • Describe what the first version really does. Do not claim features that do not exist, or results you cannot promise.
  • No fake testimonials, fake scarcity or invented numbers.
  • Disclose AI where it matters, as required by the rules where you live or by the platforms you use, and especially if a customer would reasonably want to know.
  • Be careful with regulated areas: financial, legal, medical and similar advice. Offer organisation or information, not regulated advice, unless you are qualified and permitted.
  • Reliability.

  • Test as a customer would, on a phone, with a typo, with an empty field and with a second person.
  • Have a fallback. If a tool fails, how will you still deliver? A note and a manual backup are enough.
  • Say how customers reach you. A working email or form, and a response time you can keep.
  • Back up your work and your customer list. Do not let one account hold everything.
  • Costs and limits.

  • Know what happens when you get more customers than expected. Many tools have limits and charges that grow.
  • Set a spending cap and a date to review, per your limits sheet.
  • A launch checklist.

    LAUNCH CHECK
    [ ] The one job is clear on the page
    [ ] I tested it as a customer, on a phone
    [ ] I checked who can see customer data
    [ ] I read the terms of each tool that touches customer information
    [ ] Payment uses an established tool; price, delivery and refund are stated
    [ ] No fake proof, no features that don't exist
    [ ] I have a manual fallback
    [ ] Customers know how to reach me
    [ ] My costs are inside my cap

    Step 7 — Write the scope and ship

    Now turn the thinking into a plan you can finish.

    The 14-day shape. A typical plan for a small first version:

  • Days 1 to 2: write the scope. Collect what you need: text, template, tool accounts.
  • Days 3 to 7: build the must-haves. One at a time. Test each one.
  • Days 8 to 10: test as a customer. Ask a friendly person to try it. Fix the worst problems only.
  • Days 11 to 12: write simple instructions and a short message to your first customers. Prepare the manual fallback.
  • Days 13 to 14: launch to your first few customers, or to the people who said yes in your test.
  • If your plan does not fit, cut scope rather than extend the date.

    Ship to a few people, not everyone. Give the first version to the people who signed up, booked or pre-ordered. Tell them honestly it is an early version, and ask for feedback.

    Watch, do not defend. Notice where customers get stuck, what they ask for and what they ignore. Write it down. Do not explain it away. This is the most valuable data you will get.

    Decide what goes next, using evidence. After the first customers, the next feature is the one several customers asked for, or the one that would save you the most time. Not the one you find most exciting.

    When to move to a heavier type. Move from manual to no-code when you repeat the same step often and it costs you more hours than your limits allow. Move to a coded app only when customers pay, the need is clear and you can maintain it. Each move should be funded by real income and real demand.

    A word on pivoting. If customers use your MVP differently from what you planned, take note. They may be telling you what the real job is.

    Ask AI to check your scope.

    Role: a sceptical product advisor for a first-time solo founder. You do not know my customers. Do not invent facts.
    Task: I will paste my MVP scope. (1) Tell me whether the one job is really one job. (2) Tell me which must-have could be done by hand or cut. (3) Tell me whether my ship date is realistic for my hours. (4) Name the biggest safety or honesty risk I have not covered. Be direct.
    [paste the scope]

    Worked Example: Dani's first version

    Dani, the invented meal-plan founder, ran interviews and a smoke test. The results: parents' pain is "wasted food" and "Thursday panic", and a founding offer got 7 pre-orders from 58 subscribers. Dani is tempted to build a meal-planning app with recipes, shopping lists and reminders.

    Evidence. Customers want a result: a weekly plan built around one shopping list. They did not ask for an app. Dani's hours are about five a week and Dani dislikes talking to strangers, so something that can be delivered by message or file suits.

    Choice. A mix of type 1 and type 2 with a touch of type 3: Dani will create each week's plan by hand using a template, deliver it as a printable document and send it by email through a simple email tool, with a short form for each family's preferences and allergies. No app.

    The one job. "Busy parents of school-age children get a weekly dinner plan with one shopping list that wastes less food, delivered every Sunday."

    Must-haves. (1) A short preference form; (2) the weekly plan with one shopping list; (3) a reply route for questions.

    Not now. An app, recipe database, automatic personalisation, reminders and a custom website.

    By hand. Dani chooses the meals, builds the list and checks quantities. AI helps draft meal ideas and tidy the list, and Dani checks every plan for allergens and sense. Dani also states clearly that plans are suggestions and not nutritional or medical advice.

    Safety. The form collects only family size, preferences and allergies, no names of children. Dani checks the sharing settings and the tool's terms, and tells the families how the information is used and how to ask for deletion.

    Done means. Seven pre-order families have received three consecutive weekly plans, and at least five say they would renew.

    Ship date. 12 days. Dani has five hours a week, so the plan is a template and form in the first week and the first plan sent by day 12.

    What Dani avoided. An app that would have taken more than a month and might not have been what the families wanted. After three weeks, two families ask for a "leftovers" section, so that is next. Not a login page.

    Common Mistakes to Avoid

    Three habits cause most MVP trouble.

    Mistake #1: Building software when a manual version would teach more. A beautiful tool for a result nobody has confirmed they want is wasted work. Fix: deliver by hand first, automate what you have done five times.

    Mistake #2: Letting AI generate a lot you do not understand. Code or setups you cannot read are hard to fix and may expose customers' data. Fix: keep scope tiny, ask AI to explain it back, avoid sensitive data and use established tools for login and payment.

    Mistake #3: Letting the scope grow. "Just one more feature" turns two weeks into two months. Fix: one job, three must-haves, a "not now" list and a ship date within 14 days.

    Going Further

    Try the manual version on one customer this week. Even if you plan software, deliver by hand once. You will learn more than from a month of building.

    Keep a "by hand" log. Record each step you do manually and how long it takes. The step that takes most time is your first automation candidate.

    Learn a no-code tool well. Pick one simple tool and learn it properly. Depth in one beats shallow use of five.

    What to learn next. The tutorials in this module on no-code with AI and on vibe coding are optional if you choose type 3 or 4. In the next module you will price your offer and cover the basic terms, then write your offer card in the quest.

    Key Takeaways

    In this tutorial you learned:

  • An MVP is the smallest version a real customer can use and pay for, and it is a learning tool. It does not have to be software.
  • There are four types: a manual service, a document or template, a no-code page or tool, and a small coded app. They run from lightest to heaviest.
  • Choose by evidence, process knowledge, hours and risk, using the decision path. Start with the lightest type that can deliver the result.
  • AI makes building faster and over-building easier. Use it to prepare and explain. Avoid code and setups you cannot read, and keep sensitive data out of the first version.
  • Cut to one job and at most three must-haves. Write a "not now" list, do behind-the-scenes work by hand and define "done means" with a number.
  • Check safety and honesty before launch: minimum data, established payment tools, honest claims, a manual fallback and costs inside your cap.
  • Ship to a few people within 14 days, watch what they do, and let evidence decide the next feature.
  • Before you move on, a challenge. Write the one job for your idea in one sentence, list three must-haves and at least two things you will not build, and set a ship date within 14 days. Then ask yourself: could I deliver this result by hand next week? If yes, start there.

    Learn AI, after work

    Track your progress, earn XP, and unlock more free tutorials in the AfterWork Academy app.

    Open this tutorial in the app

    More AI tutorials