AT&T Builds an AI Intake Desk for Its Lawyers With OpenAI

AT&T's in-house legal department has hundreds of lawyers and legal professionals around the world, and the first thing many of them do with a new request is hunt for documents, facts, and past work. [Fact] On October 21, 2026, OpenAI published a case study saying that this step is what AT&T wants to hand to AI first. [Fact] The piece contains quotes from two executives and no numbers at all, so the question for legal workers is which parts of the plan touch which jobs.

What AT&T says it is building

The case study describes a planned "natural language intake" system for the legal department, in the words of Bill Ryan, AT&T's Chief Compliance Officer. [Fact] Requests from across the business would enter through a single front door; the system would classify and prioritize them, connect each one to relevant precedent, research, and prior work, and hand the responsible lawyer the historical context plus recommended first steps. [Fact]

The work sits inside LegalEdge, an existing AT&T team of attorneys, legal professionals, technologists, and project managers assigned to apply AI to major legal matters. [Fact] David McAtee, AT&T's General Counsel, frames the OpenAI partnership as "plumbing" for a company-wide knowledge layer that lawyers draw from. [Fact]

The stated goal is scale. Ryan's quoted ambition is "expanding our ability to provide legal judgment at greater scale," and McAtee's stated boundary is that "the decision making will always remain with the lawyers." [Fact] Read as a task list, the plan automates gathering, sorting, and routing, and keeps advising and deciding with people. [Estimate]

What the case study leaves out

There is no headcount, no hours-saved figure, no cost, no accuracy measure, and no deployment date; the intake system is described as a planned initiative, not a running one. [Fact] The page is published by OpenAI, the vendor, and ends with a "Contact sales" link. [Fact] A single-company vendor case study of this kind shows what a buyer intends; it cannot show whether the tool changed the department's staffing or output. [Claim]

Which legal jobs do the intake work today

Classifying a request, pulling precedent, and assembling a file for the lawyer who will answer it are tasks that fall mostly on paralegals, legal secretaries, and junior in-house counsel, not on general counsel. [Claim]

Our occupation pages score task exposure with OpenAI's own "GPTs are GPTs" rubric, the same company publishing the AT&T study. For lawyers, 19 of 22 rated tasks score at or above the 0.5 exposure line, or 86.4%. [Fact] For paralegals and legal assistants it is 10 of 12 tasks, 83.3%. [Fact] For legal secretaries and administrative assistants it is 14 of 14, 100%. [Fact]

The employment outlook runs the other way from the exposure scores. BLS projects lawyers growing 4.7% from 2025 to 2035, from 864,000 to 904,000 jobs, with a 2025 median wage of $159,670. [Fact] Paralegals are projected at -0.3% over the same decade, at a median of $62,890. [Fact] Legal secretaries are projected at -5.2%, from 158,000 to 149,000, at $55,570. [Fact] The two roles whose tasks AT&T's intake system is designed to absorb are the two already projected flat or shrinking. [Estimate]

What to prepare

For in-house lawyers, the AT&T plan moves the first hour of a matter from research to review: the system arrives with context and a recommended first step, and the lawyer's job starts at judgment. [Claim] Practice in writing the one-paragraph assessment that follows a briefing, and in catching the recommended step that is wrong for this client.

For paralegals, the intake system is a knowledge base that someone has to build and maintain: matter taxonomies, precedent tagging, the rules for what counts as "relevant past work." [Claim] Those roles exist inside LegalEdge today, which lists legal professionals alongside technologists and project managers. [Fact]

For legal secretaries, the routing and prioritizing functions are the ones named explicitly in the plan, and the occupation's task list has no rated task below the exposure line on our page. [Fact] The transferable skills are matter-management software and the administrative side of compliance programs, where the work is checking that a process ran rather than running it. [Claim]

Limits

The source is a short vendor-published profile of one company, with no quantitative results, no method, and a system that is still planned; the AI exposure scores and BLS projections above come from our occupation pages and from BLS, not from AT&T or OpenAI, and the link between the two is our reading. The quotes are from executives speaking about their own program.

Sources

• OpenAI, "How AT&T is rethinking in-house legal work with OpenAI," published October 21, 2026: https://openai.com/index/att-legal
• Occupation task exposure and BLS projections as served on our pages for lawyers (23-1011), paralegals and legal assistants (23-2011), and legal secretaries and administrative assistants (43-6012).

About this analysis

This article was written with AI assistance from the full text of OpenAI's case study, read via a public archive copy after the live page returned an access block. OpenAI is quoted briefly; no images are reproduced. Occupation figures are from our own pages and from the U.S. Bureau of Labor Statistics; the mapping of AT&T's plan to occupations is ours.