CBA Record January-February 2026

AI 2035

DRAFT FASTER, NOT LOOSER A Practical Guide to AI for Transactions By Benjamin Altshul

A rtificial intelligence can accelerate transactional drafting and improve consistency, but only when it operates inside a disciplined, attorney-led workflow. The promise isn’t that an AI tool will write a better contract than a seasoned lawyer. It’s that, when confined to a closed set of approved language and guided by precise legal direction, it can propose focused revisions faster than a human can type them—while the lawyer preserves judgment, structure, and compliance with the duties of compe tence, confidentiality, supervision, and candor. The Reality of Drafting in Practice Every deal lawyer knows the scene: On one screen sits the work ing draft; on the other, a prior contract that seems close enough to adapt. Partner comments arrive as bullet-point emails or hand written notes with arrows crisscrossing the page like a treasure map, each one pointing to a secret clause you’re meant to find. Some of the handwriting borders on hieroglyphics, decipherable only after several cups of coffee and years of experience. Precedent can be uneven, especially when the “precedent” is simply the last document that crossed your desk. Meanwhile, the pressure is constant as clients expect speed and predictability, while partners expect accuracy with minimal churn. It’s tempting, in that environment, to tell an AI tool to “make this buyer-friendly” or “rewrite Section 11 like we usually do.”

That’s where trouble begins. You don’t know what data the tool is drawing from, whether its examples match your firm’s risk posture, or whether the proposed language has ever been tested under your jurisdiction’s governing law. Unconstrained AI tends to produce elegant but alien para graphs that read like brand-new provisions. Opposing counsel can see that as over-lawyering. The markup swells, momentum stalls, and the deal suffers. Broad prompts invite the tool to show how much it “knows,” often by swapping defined terms, collaps ing subsections, or rearranging sentences. The result looks tidy but becomes useless in negotiation.

The fear that AI would replace deal lawyers is fading. What remains is a divide between those who learn to work well with these tools and those who don’t.

22 January/February 2026