Prototype vs MVP vs proof of concept: what is the difference?

By Kaz Kato · Published 12 October 2026 · Updated 12 October 2026

A proof of concept shows that an idea can work at all, usually to your own team. A prototype shows what it would be like to use, to the people you want to win. A minimum viable product is the smallest version real customers use for real. Most ideas go through them in that order, and many stop early.

This guide defines each, sets them side by side with a pilot, and says which one you need and how to share each. It is written for people building with AI, who can now make a working prototype themselves. That changes what you can show a client, and how early. The prototyping use case and the page for founders show where a room fits.

What is a proof of concept?

A proof of concept is a test of whether the hard part can be done at all. It answers one technical question, such as whether the data can be matched or the model can read the documents, and it is usually ugly. It is for your own team, and it is finished when the answer is yes or no.

Nobody outside the team needs to see it. If a client must, show it on a call, where you can explain what is not there yet. A proof of concept on a link, with nobody to explain it, reads as a broken product.

What is a prototype?

A prototype shows what the product would be like to use. It looks real and does the key moments, on invented data, so a client or a prospect can click through it and say whether they want it. It does not need to be complete, secure or scalable: it needs to be convincing, and honest about being a demo.

A working prototype, one that runs rather than a set of linked pictures, is what building with AI has put within reach of people who are not engineers. It lets a prospect try the moment that matters to them, instead of being told about it. If yours began as a Claude artifact, what a Claude artifact is says who can open it.

What is an MVP?

A minimum viable product is the smallest version of the product that real customers use, with their real data, to do real work. It is built to be relied on: secure, backed up and supported. It answers the question a prototype cannot, which is whether people will keep using it, and pay for it.

An app built with AI can look finished long before it is an MVP. It becomes one when someone has made it safe for real data: who can sign in, where the data lives, and what happens when it breaks. That is usually a developer’s job. Making a prototype built with AI real walks through that step.

What is the difference?

The question each one answers. A proof of concept asks whether it can work, and a prototype whether people would want it. A pilot asks whether it works for one real customer, and an MVP whether people will use it and pay. Each asks more of the thing than the one before: more real data, and more real use.

	Proof of concept	Prototype	Pilot	MVP
Question it answers	Can it work?	Would they want it?	Does it work for one real customer?	Will people use it, and pay?
Who sees it	Your team	Clients, prospects and testers	One customer, for real	Real customers
Data	Whatever proves the point	Invented data, shaped like theirs	Real, theirs	Real
How finished	Barely	Looks real, and does the key moments	Works in their setting	The smallest whole product
Built with AI (Kaz’s view)	The quickest of the four	Quick: you can build and send it yourself	Usually needs a developer to harden it	Usually needs a developer to harden it
How to share it	On a call, inside the team	A room, with a link for each person	In their own setup	Production hosting

The “Built with AI” row is Kaz’s opinion from building this way, not a measurement.

Where does a pilot fit?

Between the prototype and the MVP. A pilot puts a working version in front of one real customer, in their own setting, with their real data, for a fixed period. It tests whether the product works for real before it is offered to everyone. A prototype wins the pilot; the pilot earns the MVP.

Because a pilot uses real data, it is not a demo, and it does not belong in a demo room. It needs the security an MVP needs, on a smaller scale.

Which one do you need?

The one that answers the question you are stuck on. If you do not know whether it can be built, make a proof of concept. If you know it can, but not whether anyone wants it, make a prototype and show it. If a customer wants it, run a pilot or build the MVP.

The order they usually come in:

1. Proof of concept — can the hard part be done at all?
2. Prototype — would the people you want to win actually want it?
3. Pilot — does it work for one of them, for real?
4. MVP — will customers keep using it, and pay?

Many ideas skip the first step, because the hard part is not technical. A sales team offering a new service, or an agency pitching a new tool, often starts at the prototype.

How do you share each one?

A proof of concept stays inside the team, shown on a call. A prototype goes to the people you want to win, on a link each, running on invented data. A pilot runs in the customer’s own setting. An MVP needs production hosting, with real sign-in, real security and someone looking after it.

A room on Roomi is built for the prototype stage. Every demo runs on mock data, and every room tells its viewers so. On Pro, each person gets a private link in their name. It can expire on a date you choose, it can be revoked at once, and you see what each person did. Sharing a private link covers the options.

When you send a link, tell the person who you are and how to reach you about their data. Say that the room records what they do there and shows it to you, as the terms ask.

If what you have is a prototype, put it in front of one person you want to win this week. Then see what they do with it.

Questions people ask

Is an app I built with AI an MVP?

Only once real customers use it, with their real data, to do real work. Until then, however finished it looks, it is a prototype: the right thing to show a client, and the wrong thing to run their business on.

Can a prototype use real data?

It should not. A prototype is shown to people outside your business, so it runs on invented data shaped like theirs. Roomi requires it: every demo runs on mock data, and every room tells its viewers so.

Is a Claude artifact a prototype?

It can be. An artifact that shows a client what a tool would be like to use is a prototype. One that only proves a calculation or a document check works is closer to a proof of concept.

What comes after an MVP?

The product itself: the MVP, extended and hardened by what its first customers did with it. From there the work is running the thing, not showing the idea.

Do clients pay for prototypes?

Kaz’s view: sometimes. A prototype made to win the work is usually yours to pay for. One made as the first stage of agreed work, to settle what to build, is often paid for as part of it.

About the author

Kaz Kato, Co-founder & CTO, Roomi. Kaz co-founded Roomi with Tesh, whom he has built with since 2017: two startups that failed, one venture studio that didn’t, and an exit.

Related guides

• I built a prototype with AI. How do I make it real?

• What is a Claude artifact, and who can see it?