
Simon Jenner
Friday 16 May 2025

AI-powered no-code platforms let non-technical founders rapidly prototype an MVP. While simple apps can be built end-to-end, complex MVPs generally reach only 60–80% completion via AI alone; the final 20%—custom features, bug fixes, scalability—still requires developer expertise or a specialist agency to finish cost-effectively.
Posted in:
No-Code
Launching a product without writing a single line of code may sound like a pipe dream for non-technical founders. But today's AI-powered no-code and low-code platforms level the playing field, enabling anyone with a vision to rapidly prototype, iterate, and validate ideas. In this post, we'll explore how viable it is to build an Minimum Viable Product (MVP) with one of these tools.
The AI dev tools
The three tools we are looking at are:
There are lots of other similar tools out there, either existing that are adding AI features or those that are built from the ground up with AI in mind. Some of the above tools have been around for quite a while like Replit and others are very new like Loveable.
All these tools promise the same dream, you can create an app by describing what you want, wait 10 minutes for the magic to happen then you have a fully working app or MVP.

Each of the tools have different ways of doing things and some require more technical expertise that the others. As you can see from the descriptions below, how they pitch themselves gives you an indication of who they are targeting.
Loveable
An AI-powered no-code platform that transforms plain-English prompts into full-stack web applications in seconds, offering drag‑and‑drop UI building and integrated backend services without writing a single line of code.
Bolt.new
An AI-driven browser IDE that generates real, editable React, Next.js, or Svelte source code from text prompts—giving you full code ownership while accelerating scaffold and deployment.
Replit
A cloud-based development environment with AI assistants, enabling non-technical users to describe an app and instantly receive a working codebase, complete with live preview, collaboration, and one‑click deployment.
Can I launch an MVP with an AI coding tool?
Yes & No, you can certainly build something that looks like a workable app from a prompt. Lets say you want to build a super simple app i.e. a loan calculator, then you could build this with a prompt and get it working no problems. However, if you want to build an MVP, lets say a patient management system for doctors and patients to be able to use then that's a different matter.
With a basic prompt it will get you about 40% of an app built. With some practice and advice on how to craft your prompt you can get to about 60% of an app build. What you can't do is get a full app built off a single prompt.
Here is what we achieved with an advanced prompt, this has a database and a sensible data structure, it has multiple screens such as one for booking an appointment.

On the face of it, this looks like a good build. However as you click around you realise nothing really works and you start to see blank pages or error messages and now the fun starts.

So can I build an MVP from a single AI prompt today = NO. Will I be able to in 6-12months time maybe in 2 years almost certainly.
Can I use AI to create the missing features?
All these tools allow you to carry on prompting via AI, so you could now add a prompt to say 'when I click Add patients I don't see a form, create me a form for new patients, the form should ask for the following input firstname (text), lastname (text), email (emai)' and the AI engine should create the code for this new form.
This way you will be able to get it much more functional, our estimate is you might be able to get it to 80% functional. The problem is sometimes AI is not that clever and it does things that make no sense and this ends up breaking the code or it gets itself into a doom loop. A doom loop is where you say this is broke fix it, it goes off makes changes, you test but the same problem exists so you ask again for it to fix it. This can go on for literally days of it trying to fix the same bug and each time its trying something new and in lots of cases creating more issues with the code.
In some cases when you get to know the tools you can take the code it produces and cut and paste it into another tool like ChatGPT or Claude and ask it to fix the code and it might be able to do it. However, as your app gets more complicated the likely hood is the code is not in one place but in lots of places and ChatGPT doesn't have the full context to understand what other code might be in play here.
So in summary you might get to somewhere between 60-80% of an MVP built using AI.
How do I complete my MVP?
Until AI improves (which it is doing weekly) the answer is you need to find an agency like Million Labs to help you complete the last 20%. The good news is that means you are only paying for that last 20% and not the full build so you can get an MVP for much less cost. If a full MVP, built in no code costs $10,000 then the last 20% is just $2,000 to get complete.
If you are stuck with your AI code build then reach out to us and we will get you unstuck.
Ready to launch your startup idea with an MVP?
Download our step by step guide for non-technical founders to create a startup Minimum Viable Product (MVP)
Get the eBook