Prototyping

Prototyping means: you first build a rough test version of an idea. This test version is called a prototype. It is deliberately unfinished, often ugly, and only partially functional. Its purpose is not to be sold, but to enable learning. The goal is to find out whether the idea holds up before much money and time flow into it. A prototype can be a paper sketch, a clicked-together screen mockup, or a program that can handle only a single task.

Why the half-finished attempt pays off

The later a mistake is discovered, the more it costs. Correcting a wrong assumption in a paper sketch takes ten minutes. Correcting the same wrong assumption in a fully programmed app can take months. Prototyping moves the discovery of these mistakes earlier. You fail early and cheaply instead of late and expensively.

The second reason is even more important: people are bad at describing what they need. If you ask ten people whether they would find a certain feature useful, most will politely say yes. If you put a prototype in front of them, you see within two minutes where they get stuck. Observation beats asking. That is why the prototype is above all a tool for generating genuine reactions.

In business, this has direct consequences. Investors today rarely want to see just a concept paper — they want something running. A working prototype proves that a team can deliver. It lowers the risk for everyone involved, because it turns a claim into something verifiable.

From paper sketch to working test model

At the start, the question is usually which assumption is the riskest. The prototype is meant to test exactly this one assumption, nothing more. Everything else may be missing or faked. A prototype of a translation app, for example, might not translate at all, but merely display a prepared text. For the question of whether the interface is understandable, that is entirely sufficient.

Experts distinguish between low-fidelity and high-fidelity prototypes. A low-fidelity prototype consists of paper, sticky notes, or simple boxes on a screen. A high-fidelity prototype already looks almost like the real product, but is empty inside. Somewhere in between lies the so-called Wizard of Oz technique: the user believes a computer is responding, when in reality a person in the next room is typing the answers. This is exactly how many voice assistants were tested before the technology for them existed.

The process is always a loop: build, test, evaluate, discard or improve. This loop is repeated several times, often on a weekly basis. What matters is the willingness to throw the prototype away. Anyone who grows attached to it and gradually expands it into the final product often ends up dragging along poor stopgap solutions. A prototype is an experimental setup, not a foundation.

Prototypes in AI products and start-up news

With AI applications, prototyping today is especially fast. Using a language model — that is, a program that generates text — a rough assistant can be put together in an afternoon. Such demos can be impressive. Still, the path from this demo to a reliable product is long, because accuracy, cost, and data protection only become real problems at that later stage.

In business news, you often encounter the term in a particular form. A start-up announces a prototype and raises money. A carmaker shows a prototype at a trade fair that never goes into series production. A chipmaker talks about first test samples from the factory. In all these cases, the same holds true: a prototype is a promise, not proof of production readiness.

A common mistake is confusing this with the Minimum Viable Product, or MVP for short. The MVP is a stripped-down but genuine first product version that already reaches customers. The prototype, by contrast, stays internal and serves only learning purposes. Anyone who confuses the two ends up shipping experimental setups to real users.