How to compare legal AI records

Start with the task, not the product

Vendors describe products. Firms buy workflows. The directory is organised by task for that reason: legal research, drafting and review, intake and client communication, practice management, document management and e-discovery, immigration, firm operations, and general-purpose assistants.

Two products in different categories are rarely substitutes. A general-purpose assistant and a case-management AI can both draft text, but only one is connected to the firm's matter data. The category on each record is the vendor's own product scope, not a CounselAtlas ranking.

Read the price path before the feature list

Every record states whether the vendor publishes a price or routes buyers to a sales conversation. That single fact decides whether a product can be evaluated without a call.

Where a vendor publishes pricing, the record links to that page. Where it does not, the record says so and links to the page that was checked. A missing price is recorded as an observation, never as a judgement about the product.

Check the firm-size scope

Firm-size bands record the vendor's own language. An explicit solo-practitioner statement maps to Solo, small-firm language to 2-5, and mid-sized-firm language to 6-20.

When the checked source does not state a firm size, the record says Not stated and the product is omitted from firm-size-filtered results. That keeps the filter honest rather than inferring a fit.

Then read the confidentiality panel

The confidentiality panel quotes the vendor's own wording about client data, with a link to the page it came from. Where no qualifying statement was located, the record says Not stated.

Not stated is not an accusation. It means the checked source did not contain the statement. Absence of a published commitment is recorded as absence, never converted into a promise or a warning.

Sources

• CounselAtlas records observations with a date and a source link. Open ↗