There’s a pattern I’ve come to recognize with almost every generative AI tool I use: it gets me 80% of the way there, almost instantly. Whether it’s drafting an email or generating code, the early output has something magical about it. Yet, that other 20% is where meaningful effort becomes essential. The process of fine-tuning the result through quick tweaks often feels like the futile exercise of pushing a rope. The initial magic wears off to expose the need to bring one’s own knowledge to bear in order to successfully arrive at the final target.
This personal experience aligns closely with my own experiences within my professional practice. I work at an AI consultancy that helps businesses navigate the complexities of artificial intelligence. A common scenario is managers or business owners coming to us with a broad question: “What can we do with AI?” They’ve heard the buzz and are eager to apply what they heard.
However, after we guide them through a design sprint to clearly define their requirements, a funny thing often happens. The problem they experience rarely requires an advanced form of an AI model at its core. Instead, what is really needed is a more streamlined business process. As Cedric Muchall (2022) points out in his book Bedrijf Bamischijf, companies often fall into the trap of adopting trendy solutions instead of addressing the fundamental pitfalls of how they are organized.
AI tools can enhance a step in that new process, like summarizing reports or automating data entry, but it’s a supporting actor, not the main star. The core solution is almost always about rethinking the workflow, not just plugging in an AI. The technology becomes a small part of a much larger, more human-centric solution.
For me, this highlights a critical misunderstanding in the corporate space. The temptation is to look for a technological fix, but the real gains come from first understanding the fundamental problem. It seems the hardest part isn’t the AI implementation, but the human process of figuring out what to build in the first place.
Muchall, C., & Toma, L. (2022). Bedrijf Bamischijf: van onzinnig “bedrijfje spelen” naar zinnig organiseren.
I enjoyed reading your post a lot! It got me thinking, maybe the boost from ai is less about the 80%, but more about how it delivers that 80% with great confidence. It looks almost finished which makes people lazy, but under the hood there is a lot of unfinished work and thinking still to be done. By presenting a product that looks almost finished, people’s natural scepticism is lowered, resulting in less effort to look for the underlying problems. On the other side, when a colleague proposes a great idea, that is not finished this can lead to debates which take time and effort. But when AI presents a moderately good idea that looks nice, people may choose to just roll with it.