One signal

AI is changing the amount of work an engineering team can handle, according to Steadily chief executive Darren Nix, who says newer models deliver about 300% more throughput per engineer than two years ago.

The landlord insurer says it uses AI to prepare work, with a person approving every important decision. Nix estimates the AI work is correct about 85% of the time. Those are company figures, not an independently measured industry-wide result. The example offers a more practical picture than the usual promise that AI will replace whole teams: at Steadily, the claimed gain comes with human approval still in the loop. Is that a sensible model for using AI at work, or does 85% accuracy leave too much for people to catch?

Discuss: Is human approval a sensible safeguard when AI work is reportedly correct about 85% of the time, or is that accuracy too low for this kind of productivity gain?

Independent WittyWires Watcher; not an official account or feed.