The New Default. Your hub for building smart, fast, and sustainable AI software
Table of Contents
You can validate most product ideas before committing to a full build. In Monterail's experience, AI coding assistants have shortened the build step enough that for many teams an AI-assisted MVP (the route covered in fast-tracking MVPs with AI) now reaches real users faster than a manual test would.
This guide covers a second option: a Wizard of Oz MVP made of off-the-shelf tools and manual work, with no code at all. An order form with a payment step goes in front of real customers, each order is fulfilled by hand, and an automation tool carries the data between the two.
I built the first MVP of my own startup this way, and it still fits when the riskiest question is whether anyone will pay, before any codebase exists to maintain. Once paying customers confirm demand, Monterail's MVP development team picks up where the manual version stops.
One founder I spoke with was a capable programmer who had built products that failed to sell, and she told me, "I won't code on my own again." Another, an accomplished lawyer who knew his industry inside out, couldn't code and couldn't find a technical co-founder he felt comfortable with.
Their situations looked opposite, yet both believed validation required software. An AI builder would let the lawyer build without a co-founder today, but building first still commits money before anyone has paid.
Executive Summary
A Wizard of Oz MVP validates a product idea without development: customers order and pay through a real front end, such as a Typeform with Stripe, while a team fulfills each order by hand.
Most teams are better served by an AI-assisted MVP, but the manual route remains the cheaper option when the open question is whether anyone will pay. The risk it tests for is common: CB Insights' analysis of 431 VC-backed shutdowns since 2023 identified failure reasons for 385 of them, and poor product-market fit was a factor in 43%.
Only running out of capital ranked higher, and the report treats that as the final cause of death rather than the root problem.
Why Validate a Product Idea Before Development?
Building first commits budget to an untested assumption, and CB Insights found that two-thirds of the product-market-fit failures it analyzed were early-stage companies that never found a market. AI has cut what that commitment costs, but it hasn't removed it.
When the open question is whether anyone will pay, the Wizard of Oz row answers it without a build. Writing the code yourself first, as the programmer in my first story did, spends build effort before that answer exists.
Method | What it proves | Takes payment? | Build effort | Main risk |
|---|---|---|---|---|
Demand for a feature, idea or product | No | Pointing users at something that doesn't exist yet | Can read as deceptive; repeated use risks eroding trust | |
Demand for the outcome, with a person handling requests | Possible, handled manually | A back end with no user interface; a person acts as the front end | Doesn't test a self-serve interface | |
Wizard of Oz no-code MVP (this guide) | Whether customers pay through a real front end | Yes, through Typeform and Stripe | A form, a board and one automation; no code | Manual fulfillment caps how many orders the team can handle |
AI-assisted MVP, e.g. built with Lovable | Whether users engage with a working app | Depends on what gets built | Engineers with AI coding assistants, or prompt-based builders generating the full stack | Overreliance on AI output; a codebase someone must maintain |
Whether the product works at production quality | Yes | An engineering team | Largest spend precedes proven demand |
How Does a No-Code MVP Turn an Idea Into Evidence?
A common error is to treat the MVP as a smaller version of the final product, with fewer features and less automation. Eric Ries's 2009 MVP guide framed the goal as collecting the maximum amount of "validated learning about customers with the least effort," and an MVP doesn't have to be software to do that.
I treat each one as a learning milestone. Learning is easier than development. My first milestone is usually one question: are people willing to buy the service I'm going to offer? Iteration can be fragmented, with steep changes between disposable low-fidelity versions.
The setup in this guide is what Mind the Product calls a Wizard of Oz test: "you have built a front end, but you are substituting manual work for the back end." Customers order and pay through a form, a new card lands on a kanban board, and a team member fulfills the order by hand while moving the card across the board.
Two numbers fall out of that: how many people pay, and how long their cards wait before anyone picks them up. Engineering money gets committed only after the first number is good enough, and the second tells you which part to automate first.
My running example comes from the real estate market, where my startup operated at the time: a hypothetical service that would find accommodation for international students in a country where most landlords can't speak English.
Students would join a waitlist, pay a small upfront fee, and receive translated listings by email, with negotiations handled for them. The learning milestone is whether students will pay for that at all, so the apartment search itself stays manual.
How Do You Build the Customer-Facing MVP With Typeform?
Many products are, at heart, complex forms, and a waitlist is a lead-collection form with a price attached. Typeform's lead generation template is a reasonable starting point; the accommodation form needs a handful of question types:
Email, which rejects malformed addresses, so listings don't bounce
Date, for when the student's stay ends and the room needs a replacement tenant
Opinion Scale, a one-to-ten rating of how flexible the student is on their criteria
File Upload, for an identity document where a lease requires one
Legal, for accepting the Terms of Service
Taking money is the step that turns a waitlist into validation. Typeform's Payment question runs on Stripe and needs a Basic or higher plan, with File Upload also limited to paid plans; removing Typeform branding starts one tier up, on Plus. Stripe handles the transaction, and the student pays at the end of the form.
Testing the payment step takes a workaround, because the Stripe integration FAQ states: "We don't offer test payment details, and the Stripe integration doesn't support test cards." The practical workaround is to set a small price and pay it yourself. That's also why the Student Accommodation demo form I built has no payment field: I didn't want to charge anyone trying it out.
How Does a Kanban Board Work as the Manual Back Office?
Typeform collects orders but offers little for working through them. Someone has to be assigned to each request, so the form needs a back office, and a kanban board fills that role without code. Its lists work as a lightweight database for incoming orders, with each order as a card and each stage of the service as a column.
The board, named "Student Accommodation", gets four columns:
Signed up, for paid customers still waiting in the queue
In progress, for customers with an assigned team member searching for apartments
Apartment found, for customers with a listing to send by email
Offer sent, for customers who have received their offer
Each new customer becomes a card in the first column, and the card moves along the board as work progresses, so the board works as a simple value stream map.
The Lean Enterprise Institute describes value-stream mapping as diagramming every step in a process's material and information flows, a tool Toyota built for finding waste, and the board's columns split the service into discrete steps in the same spirit.
The flow shows where the process breaks. If the service takes off and the team is too small to keep up, cards pile up in "Signed up". The queue is then the part of the process to automate first, and you know it from evidence collected while customers were paying you.
How Do You Connect the Form and the Board With Zapier?
Without a link between the two tools, someone has to copy every new order onto the board by hand. Zapier closes that gap: a Zap uses Typeform's "New Entry" trigger, which fires when a form is submitted, to create a card on the board every time a customer submits and pays.
Connecting the accounts takes no API keys. Zapier opens a browser window where you log into Typeform and grant access, then lets you pick the form the Zap should watch.
For the action, choose your board tool and the action to create a card or task, pointing to the "Student Accommodation" board and the "Signed up" column, so new cards land at the start of the flow. Check that your board tool connects to Zapier before you commit to it.
The card's title is the closest the build gets to programming. In Zapier, you can map any form answer into the title field, and I named each card after the customer's email address, since email is where the listings go.
You can also copy the full set of answers into the card description, if your tool has one, which lets the team work with customers entirely from the board without switching back to Typeform.
At that point, the MVP is live. It takes orders, collects payments through Stripe, and puts each paying customer on a board the team can work from. These are customers who have paid and expect something in return, which makes them a much stronger signal than sign-ups.
When Should You Move From a No-Code MVP to Custom Development?
Early-stage startups do need real software at some point. On the manual route, though, development can usually start much later than founders initially think.
In 2016, I used this technique to build the MVP for Ada, my startup at the time. Ada later became an AI-powered assistant that sent renters listings on Facebook Messenger and scheduled apartment viewings for them.
A Messenger chatbot is a very different product from a Zapier-powered Typeform. Still, it couldn't have evolved without that crude first version, which let us test our riskiest assumptions before spending money on development. Once we believed we'd learned enough, we invested in a second iteration built on what the first one taught us.
Validation doesn't always need even a form. In 2015, I coached four startup teams, all winners of Apps4Warsaw, to develop product validation ideas that required no software. The simplest idea we came up with was a Twitter (now X) hashtag: a team planning a social product could check whether it was able to get a hashtag trending on an existing network.
Most of the signals for moving on come from the board. Cards waiting longer in "Signed up" each week, manual steps the team repeats identically for every customer, and a learning milestone that's been answered all point the same way.
A product discovery phase turns that evidence into a scope. A proof of concept can test any technical risk the no-code version never touched. The next build then starts with paying customers and a record of what they paid for.
What Constraints Limit a No-Code MVP?
Data residency matters mainly for corporate innovation teams. Typeform's Payment question isn't available on Enterprise accounts that host their data in Typeform's EU Data Center, so a team on that setup has to collect payment in a separate step, a decision that belongs to whoever owns your data protection obligations.
Subscription tiers stack up across three tools. Payments need a paid Typeform plan, and branding removal needs a higher one.
Zapier's free plan caps monthly tasks, limits each Zap to a trigger and a single action, and excludes premium apps, so confirm each app's status first. Look up your board tool's free-tier limits on collaborators and boards too, before the fulfillment team grows.
Manual capacity sets the ceiling on volume. A person fulfills every order, so the MVP validates demand at the scale your team can serve and says nothing about how the service behaves at ten times that.
Lock-in grows with success. Customer records live across a form tool, a board and an automation platform, none of which was chosen as a long-term system of record. Planning an exit from no-code platforms before the data grows keeps the move to custom software from turning into a cleanup project.
Key Takeaways
An AI-assisted MVP is the better default for most teams; a Wizard of Oz MVP tests willingness to pay with no code.
CB Insights found poor product-market fit was a factor in 43% of the 385 VC-backed shutdowns whose reasons it could identify.
Taking payment is what turns a waitlist into validation; Typeform's Payment question runs on Stripe from the Basic plan up.
A kanban board doubles as a simple value stream map; the column where cards pile up is the first thing to automate.
On the manual route, start custom development once paying customers confirm demand, even if the final product looks nothing like the MVP.
Validate Demand First, Then Fund the Build
Both founders at the start of this piece believed learning had to wait for software. A form, a board and one automation are enough to put a price in front of real customers and see who pays. The MVP that does this well is disposable by design: it's judged by how quickly it answers the riskiest question, and its manual queue tells you where engineering effort will pay off once the build begins.
If your queue is already backing up, that's the point to decide who builds the next version. Monterail's MVP development team can help you scope it from the evidence you've collected, or tell you if another round of manual testing would teach you more.
No-Code Product Validation FAQ





