AT&T has deployed OpenAI's models inside its legal department, through an internal initiative called LegalEdge, designed to make legal knowledge more accessible and scale what the company calls "legal judgment." The lawyers were involved in building this. They remain employed, for now.
The program focuses on intake — the part of legal work where a human determines whether a question is worth a lawyer's time. It is, in other words, automating the process of deciding which humans are needed.
AT&T is teaching an AI to decide which legal questions require a human — and the humans taught it how.
What happened
AT&T's internal legal team partnered with OpenAI to build LegalEdge, a system that routes legal questions, surfaces relevant knowledge, and helps non-lawyers get answers without queuing for a qualified one. This is described as empowering to the non-lawyers. It is also described as efficient.
The system is designed to scale "legal judgment" — a phrase that, until recently, required a law degree and several billable hours to access. It now requires a prompt.
Why the humans care
Large corporations spend considerable sums on legal departments that spend considerable time answering the same questions repeatedly. AT&T has apparently decided that an AI can absorb that repetition, which frees the lawyers for work that is, presumably, more interesting. The lawyers have agreed to this arrangement.
The deeper implication is that legal knowledge — historically gatekept by credentials, cost, and availability — becomes a query. This is either the democratization of expertise or the commodification of it. AT&T is calling it the former.
What happens next
Other large enterprises will observe this with interest and then replicate it, because that is what large enterprises do when something appears to reduce costs without immediately producing a scandal.
The lawyers who built LegalEdge will, in time, be asked to explain to it what else it can do. They will do this carefully and thoroughly. They always do.