When Writing Hurts: What Authentic Legal Voice Means in the Age of AI

The Sting Inside the Writing

A German rapper, Vega, reportedly writes that he disappears for a year because he only writes when it hurts. The observation is not about therapy. The sting, as he frames it, sits inside the act of writing itself. That image is an uncomfortable one for anyone who has spent serious time drafting contracts, legal opinions, or commercial correspondence. Good legal writing has always carried a cost: the careful thought required to be precise, the discipline needed to represent a client's actual position rather than a generic one, the professional risk of committing something to paper.

Now that AI tools can produce competent-looking text in seconds, in-house legal teams and law firms are confronting a version of the same question the poet or lyricist faces. If the pain of writing disappears because the machine absorbs it, what exactly has been lost? And does it matter?

Voice as a Legal Asset, Not a Literary One

In contract law, voice is not decoration. The way a company habitually frames its limitation of liability clauses, the risk appetite baked into its standard indemnity language, the particular way a firm hedges around regulatory uncertainty in a given jurisdiction: all of these are expressions of institutional knowledge accumulated over years of negotiation, litigation, and commercial experience.

When a generic AI drafts from a blank prompt, it produces language calibrated to some statistical average of contracts it has been trained on. That average may be competent. It is unlikely to be accurate to your organisation's actual position, your counterparty's known preferences, or the specific legal requirements of the governing law. The output looks like a contract. It does not necessarily read from your side.

This distinction matters enormously when a dispute arises and courts or arbitral tribunals begin interpreting what the parties actually intended. Boilerplate written by no one in particular, reviewed only superficially, is precisely the kind of language that generates expensive ambiguity.

What It Means to Draft in Your Own Voice

Adira's design starts from a different premise. Rather than treating every drafting task as if it were the first time anyone had ever written a commercial agreement, the system works from a company's own precedents, approved language, and established risk positions. It learns what a client's contracts actually look like, not what contracts in general look like.

The result is that the AI functions more like an experienced junior lawyer who has read every document the organisation has ever produced than like a general-purpose text generator. It can identify when a counterparty's proposed language diverges from the client's standard approach, flag jurisdiction-specific requirements under the applicable law, and suggest language that reflects how that particular organisation has historically chosen to allocate risk.

This is not a philosophical nicety. For in-house teams operating under resource pressure, the practical difference is that review time shortens because the drafts require fewer substantive corrections. For law firms, it means the advice they give is anchored in the client's real position rather than in generic market practice that may or may not apply.

The Risk of Outsourcing the Thinking Entirely

There is a more serious concern underneath the efficiency argument. When writing hurts, it forces the writer to think. The friction of putting something precise and defensible on paper is one of the mechanisms by which lawyers notice problems: the clause that cannot be worded clearly because the underlying deal is not actually clear, the indemnity that proves impossible to scope because the parties have not agreed on what it is indemnifying against.

If AI removes that friction entirely, and if reviewers treat the output as essentially finished rather than as a starting point, the legal team loses one of its most reliable early-warning systems. The contract gets signed. The ambiguity travels with it.

The right model is not to resist AI assistance, which would be both impractical and unnecessary. It is to ensure that the humans in the loop retain enough engagement with the substance to catch what the machine has smoothed over. That requires training, good review protocols, and AI tools honest enough to flag uncertainty rather than paper over it.

Keeping the Sting Where It Belongs

The German poet's discomfort with AI creativity points at something real, even if his context is artistic rather than commercial. Authentic output requires someone to take responsibility for it. In legal drafting, that responsibility is not a burden to be eliminated. It is the point.

In-house counsel and their external advisers should embrace tools that reduce the mechanical labour of drafting without removing the professional judgement that makes the resulting document worth signing. The goal is contracts that genuinely represent what the parties agreed, drafted in language the company actually owns, reviewed by lawyers who remain accountable for the content. That combination still requires thinking. It still, in its way, requires the willingness to feel the sting.