AT&T automates legal intake with OpenAI to free lawyers for strategy

The story doesn't clarify whether AT&T built custom legal models or is licensing OpenAI's base models with domain fine-tuning. That distinction matters: owning the model means AT&T controls liability and competitive moat; licensing means OpenAI captures margin and AT&T remains dependent on a vendor.

This is the second major legal services deployment in three days. Thomson Reuters embedded CoCounsel into ChatGPT Enterprise on October 19, betting that lawyers need citation-backed, auditable reasoning. AT&T's move two days later suggests the legal market is fragmenting between platform plays (Thomson Reuters layering onto OpenAI's infrastructure) and direct integrations (AT&T building in-house workflows). The pattern mirrors what we saw with Shield and Uber: large enterprises are moving past pilots into operational embedding, but the question of whether they build, integrate, or license remains unsettled. If AT&T is licensing base models while Thomson Reuters is selling domain-specific tooling on top of them, the market is stratifying by expertise layer rather than consolidating around a single vendor.

Within 90 days, watch whether AT&T publishes case metrics (matters triaged per lawyer, time saved, error rates) and whether those numbers are comparable to Thomson Reuters' benchmarks. If AT&T's internal build outperforms licensed solutions on speed or cost, it signals enterprises will invest in proprietary legal AI stacks. If Thomson Reuters' auditable reasoning becomes table-stakes for regulatory acceptance, AT&T will need to retrofit that capability or lose competitive ground to firms using CoCounsel.