AI in digital projects: What we learn from every MVP
Not long ago, AI was the hot topic in strategy meetings and nobody really knew what to do with it.
Today? AI is not just here - it is changing how we approach digital projects. It is changing how we build digital products - from planning to go-live.
PoCs are back - and more important than ever
Were you there when desktop applications were replaced by web applications? Then you probably experienced how often we had to prove that it worked at all: That a web app can be just as performant, secure and reliable as its desktop version. Today, nobody questions this anymore - it's simply standard. With AI, we are back at this point: we first have to show that it works not only in theory, but also in practice.
PoCs are essential. Testing. Experimenting. Validate. Can the model deliver the quality we need? Does it scale? How does it affect UX and business processes? The answers are only available if we build and test it.
Testing becomes the supreme discipline
Classic software is predictable: Input in, output out, done. And yet bugs are part of the daily routine and are kept in check by technical and methodical measures in modern software development processes.
AI-supported systems? Quite differently. They work with probabilities, learn and change.
This means:
❌ Fehler sind nicht mehr binär – sie sind graduell.
❌ Performance kann schwanken.
❌ Was gestern funktionierte, kann heute scheitern. Trotz gleichem Input.
❌ Bias, Schwankungen, unerwartete Ergebnisse – sie gehören dazu.
Companies implementing AI projects need a new testing culture. Traditional approaches are no longer enough. Instead, monitoring, feedback loops and continuous fine-tuning are coming to the fore.
Why this is crucial for digital strategies
For digital strategists, CTOs and innovation leaders, the question is not whether AI is relevant - but how to integrate it into existing digital projects in a meaningful way. Here are the three most important lessons learned from our AI projects to date:
- Experimentation is a must. Anyone looking for the "final solution" from the outset will fail. Instead: Small prototypes, iterative testing, learning.
- UX decides. AI is not just a tech topic. If users don't trust it or the interaction is incomprehensible, even the best AI is useless.
- Think long-term. AI models are not fixed systems - they need to be maintained, tuned and adapted to user feedback in order to get the most out of them.
AI doesn't just change software - it changes it, how we think about digital projects. Unternehmen, die das verstehen und sich anpassen, haben langfristig die Nase vorn.
Ob AI beim Schreiben dieses Artikels geholfen hat? Klar. Und du? Lebst du noch im 2023? 😉
Written by
Josh Wirth
