AI Document Review vs Associates: The Wrong Comparison

The comparison gets framed as a swap: pay an associate to review documents, or pay software to do it. Put that way, the answer looks like a spreadsheet. It is the wrong frame, and firms that adopt it tend to buy badly.

Associates do not just review

An associate reading a file is also learning the practice, forming judgment about what matters, and building the context that makes them useful in three years. Remove the reading entirely and you have optimised this year's throughput by mortgaging your bench.

The reviewable work splits cleanly

Separating a scanned block into documents, labelling them by type, ordering them by date, removing duplicates, extracting dates and provider names: this is mechanical. Deciding what is material, what is inconsistent, and what to do about it is not.

• Mechanical: splitting, classifying, ordering, de-duplicating, extracting.
• Judgment: materiality, credibility, strategy, and what to concede.
• The first is worth automating. The second is the practice of law.

What changes for the associate

They still read the file. They read an organised one, and they spend their time on the questions that build judgment instead of on page separation. That is a better training year, not a worse one.

The professional obligation does not move

Whatever produced the first pass, a licensed attorney is responsible for what leaves the firm. Any process that obscures that, or that makes checking harder than doing the work again, has not saved anything.