My interest in AI began around 2020, when discussions about Artificial Intelligence were mostly focused on research and early applications, especially in Europe, long before generative models were widely available. At the time, AI was mostly a technological curiosity, rather than a practical tool for legal work. I did not know where AI could first make an impact, but it felt essential to understand the underlying technology early. 

Having spent more than two decades working across computer science technologies, telecommunications, and automotive systems; first as an R&D engineer and later as a European and French patent attorney, I had witnessed multiple technological shifts that rapidly transformed industrial innovation.

So I decided to pursue formal training at CEIPI’s AI & IP diploma, and later continued with data science courses, simply to be prepared for what might come. What I did not anticipate was how quickly AI would become a practical tool for patent professionals.

Early experiments: interesting, but not transformative

My first encounters with drafting support based on AI (around 2021) was far from convincing. At that time, most models available to the public were general-purpose language models with little understanding of patent-specific vocabulary. This was also a period in which the underlying AI models themselves were far less capable than those available today; before large-scale, long-context generative models became practical enough for professional use. In most cases, it was still faster (and safer) to draft using classical methods.

However, these early trials revealed something important: if AI was ever to become genuinely useful in our field, it would have to be far more than a text generator. It would have to understand patent reasoning, not just produce language.

The real shift: AI native to the drafting process

Over the past year, I have seen a notable evolution, particularly as large language models have developed at an extraordinary pace. AI tools have become less like text generators and more like native drafting companions. Rather than producing complete drafts, they now assist in structuring the description, aligning embodiments, and checking consistency across sections.

In other words, the drafting process itself is becoming AI-enabled. This applies whether the invention concerns a telecommunication device or method, a computer-implemented method, an automotive component, a mechanical device, an AI based invention, etc. The subject matter does not matter: the drafting reasoning does.

What is emerging today is less a one-click drafting tool and more a collaborative workflow in which the attorney remains firmly in control. AI can be seen as a super assistant that is driven by the attorney. AI can support structure and help maintain internal coherence during drafting, checking consistency with technical terms used, proposing academic definitions, proposing supplementary advantages for the invention.

Crucially, these AI systems do not replace legal reasoning or examiner interaction; their value lies in reinforcing consistency and improving the analytical foundation of our work.

The future

Just like how digital prior-art search once transformed the search phase, and how electronic filing transformed administrative tasks; AI-native drafting support is beginning to transform how we write, analyze, and refine patent applications.

I do not believe AI will draft patents autonomously. I do believe AI will become a normal element of patent practice - used by experienced professionals to elevate quality, reduce inconsistency, and enable more analytical, more strategic drafting.

This will not reduce the role of the patent attorney. On the contrary: it increases the importance of skilled reasoning. As patent professionals, we will spend less time re-formatting language, copy-pasting claims, checking the references of the different elements described etc., and more time defending inventive contribution and ensuring legal robustness.

The result is not automation. It is augmentation. A gradual evolution of the profession itself.