Who will be the chefs in the AI kitchen?
In his legal column, Marijn wonders: who takes responsibility and prevents us all from falling victim to poorly configured and badly managed intelligent legal systems?
Honestly? It gets tiring sometimes. Since ChatGPT, it feels like the world is moving in fast-forward. Are you experiencing this too? Pilots, experiments, and launches everywhere. My LinkedIn timeline is nothing but AI, filled with one expert after another. The expectations of legal professionals are sky-high, and many around me are suffering from a severe case of FOMO. It makes people restless: AI is going to speed up legal processes, automate standard work, and maybe even fundamentally change our profession, or make some of us redundant.
What isn't new
We are in the midst of the AI hype, but it is good to realize that technology in the legal sector did not start with AI. For over ten years, we have been experimenting with document automation, e-discovery, and contract lifecycle management systems. It was all slow, fragmented, and to be honest, in many cases, those tools were barely used.
The difference now is the speed and the massive scale. Expectations are higher than ever. Where a legal team used to have full control over if and when they would adopt new technology, that autonomy is now under significant pressure. Not only because technology is developing faster, but also because not using AI is increasingly met with incomprehension from clients and internal stakeholders.
From tool to core process
Furthermore, the nature of the technology is changing. Previously, legal tech was primarily secondary—a helpful add-on, a tool. The reaction of many lawyers to AI is still in line with that mindset, but those who actually work with it know better. The next generation of legal technology is developing into an integral part of legal work. It will soon determine not just how fast you deliver, but also the quality and reliability of your output, right down to the service level clients experience. No, of course that is not fully reality yet, but it is arriving at a rapid pace.
Compared to the past, one thing remains unchanged: if you think you are there by simply purchasing a few licenses, think again. Anyone who has ever bought legal tech knows that the risk of disappointment is higher than the chance of success. Who hasn’t heard stories about disastrous CLM implementations, or contract drafting tools used by almost no one because everyone kept copying and pasting into their own templates?
Of course there are bad tools, but in most cases, it is not the tool's fault. Disappointment is almost always traceable to a lack of genuine implementation, defined processes, and clear ownership. Combine that with poorly managed expectations, and if results aren't immediate, users disengage, creating a negative atmosphere around the technology. It’s like buying an expensive kitchen and expecting Michelin-star dinners to appear on their own. Everyone understands that doesn't work. Why should it be any different with legal tech?
Different skills
To stay with the kitchen metaphor, at the very least it requires a chef who knows exactly how to operate all the equipment. And if it’s a large kitchen, perhaps an entire kitchen staff. With legal tech—and AI is no different—it works the same way. Being a good lawyer is no longer enough. A legal team with technology integrated into its core processes—which will eventually apply to everyone—must also be able to configure and audit its own systems, set up processes, ensure AI governance, and analyze data. This applies not only to the systems legal uses internally, but also to systems that enable non-lawyers elsewhere in the organization to perform part of the legal work themselves.
This configuration, oversight, and monitoring is currently often done by lawyers as an