Put AI in the Middle of the Search. Keep It Away From the Ends.

The first wave of AI in recruiting was meant to read every CV so nobody else had to. It worked for about a year.

Then candidates got the same tools. CVs began mirroring job descriptions word for word. Cover letters became fluent, specific and indistinguishable. A screening model trained to reward keyword match now rewards whoever prompted best. Application volume went up, signal went down, and recruiters ended up reading more, not less.

So the fully automated search failed. Firms getting results now use a split known as 10-80-10.