Experience with AI

The Use of AI in Arbitration: What Happens When Both Sides Largely Leave Their Submissions to a Machine?

Fact is: in large law firms, AI has become standard in commercial disputes, not the exception. Harvey, the dominant legal AI platform, states that 80% of the 100 highest-grossing US law firms use AI. Among smaller firms and sole practitioners, the picture is probably far more varied, up to an outright refusal to use AI.

The corresponding error typology is documented for public proceedings: fabricated citations, misrepresented genuine decisions, false quotations. These errors are extremely dangerous because the source does not support the statement. Little of this leaks out of confidential arbitrations. But once the tribunal notices AI-based errors, the entire submission is discredited – that, for me, is a given.

For an experienced practitioner, however, AI is recognisable, or at least palpable: uniform grammar, formulaic transitions, recurring rhetorical patterns, great length with little argumentative depth, a mirror-image structure. And, of course, the “ping-pong argument”: one side makes a point, the other responds at equal length – the submissions balloon, and both sides drift further and further away from the heart of the matter. A particularly “fine” example from my practice:

Both law firms had evidently made extensive use of AI, recognisable by soulless diction, little engagement and unmistakable ping-pong.

I like to follow my favourite principle: “There’s always a story behind the story.” Something about the case had a certain “haut goût” to it; as a US colleague put it: “The matter stinks.” The arguments got lost in ever smaller details and strayed too far from the core. Then I had an idea that had evidently not been part of any AI prompt, and which the AI, in the context of the claim, had not “thought” of on its own: I looked into how the claimant company presented itself publicly and, alongside it, the self-portrayal of its managing director.

“Bingo”: on the claimant’s own website, in a hidden drop-down menu, I found a talk the managing director had given to an industry association. In it, he stated that they had designed the large plant themselves – “building the plane while flying it” – and that, lo and behold, their own team had overestimated and overstretched itself technically as general contractor.

Anyone who describes themselves this way in public will hardly be able to shift responsibility for the plant onto someone else, as was attempted in the proceedings. Yet this admission was nowhere to be found in the submissions. Instead, caught up in their ping-pong, the law firms lost themselves in remote areas of law …

We all know that AI output has to be checked. My point is a different one: intuition, a “nose” for things and our experience as practitioners cannot be replaced by AI. Humans beat the machine with their intuition. We have to stay on the ball and must not allow ourselves to be replaced.

So much for the potential value of a second opinion in ongoing proceedings.