The AI Exposure Index: Law

What does this Index measure?

Every competitor in this census is selling today. Some sell software, and some sell the finished work the way a law firm does. What decides a practice’s exposure is whether the client can get that finished work without the firm. Where the client can, we mark the practice replaced. Where the client would still need a licensed lawyer it doesn’t have in-house, the firm keeps the matter at fewer hours, so we mark the practice repriced. Open a practice below to see the companies attacking the work it bills for.

Where did the companies come from?

The field was built by systematic web research: one sweep per legal practice, plus a completeness pass for what the sweeps missed. A company earns a place only with a sourced account of what it sells; anything we could not tie to a primary source was dropped, not estimated. That produced 221 evidenced entries, which resolve to 162 distinct companies. One company often sells into several practices, and a product counts under the company that owns it, so Harvey and Harvey for eDiscovery are one, and CoCounsel, Westlaw AI and Materia are all Thomson Reuters. A further 3 researched companies came out under the exclusions below.

How was each company placed?

A company is placed on a practice when it targets the work that practice bills for, and the verdict sits on the practice rather than on the company. Horizontal platforms appear on every practice they sell into, described at the product level rather than practice by practice, so the same summary can appear under several practices below. The Index carries no tag for who buys a product: a vendor’s marketing names who it courts, not who can buy.

What counts as a practice?

A company is attached to a practice only when it targets the work that practice bills for, not a slice of scaffolding under it. That rule took three practices out of the Index entirely, once their apparent competitors turned out to be compliance or document-review tools aimed at adjacent work. What remains is 21 practices and 168 placements, naming 124 of the 162 companies. The other 38 compete for work this Index does not yet carry a line for. They are mostly consumer-facing practices this Index has not censused yet, family law and divorce, employee-side employment, consumer and humanitarian immigration, and real estate and title, plus the research-and-drafting layer that sits under every practice rather than competing for one. That is a limit on this Index’s coverage, not evidence those practices are safe: the buyer-side pressure on them is if anything heavier than on the practices below.

What does this Index not cover?

Back-office tooling (billing, intake, practice management, docket analytics) is excluded as adjacent to the firm, not a threat to a practice: an AI receptionist and an invoice-review platform are out, while a product that drafts the billed work stays in even when it also handles intake. Work that is not legal practice is excluded on the same test, which takes out security and financial-compliance automation, and tax-return preparation sold to accounting firms. Some categories are still under-covered by the sweeps, so treat the counts as a floor, not a ceiling. The Index also does not weight practices by their share of a firm’s book, because that weighting would be editorial judgment, not a sourced figure.

How should I read it?

A competitor named here isn’t lost revenue. We read shipped product and live service as of September 2026, so the Index shows who is coming for the work, not billing that has already moved.

Sources

• Armour, Parnham & Sako, Augmented Lawyering (University of Illinois Law Review, 2022)

  The augment-versus-disintermediate distinction this Index is built on.

• Clio, Legal Trends Report 2025

  Adoption running ahead of repricing, the lag described above.

• Briggs & Kodnani, The Potentially Large Effects of Artificial Intelligence on Economic Growth (Goldman Sachs, 2023)

  Legal's task-level exposure, and why most of it lands as complement rather than substitution.

Every company named in the Index links to its own site, so each entry can be checked at the source.

Download the dataset

Reviewed October 20, 2026. Licensed CC BY-ND 4.0. Share and cite it with attribution and a link back, but don’t republish a modified or rebranded copy.

JSON · CSV

Corrections and additions

This is a living census, not a closed report. If a company is missing, sits under the wrong practice, carries a verdict you can argue with, or has a better primary source, send it with a link. Useful for anyone who sees a gap in a practice they know, and for vendors who want the record accurate.

shawn@upshiftco.com