Voices
Benedict Evans: the bottleneck on AI at work is recognition, discovery and institutionalization, not model capability
In "AI, Tools and Transformation" (September 3, on the HN front page today), Evans argues the automate-your-own-job thesis fails on three counts: a great matrimonial lawyer does not think about their work as an engineering problem, successful automation targets were historically not obvious even to technologists, and anything touching "50 or 500 people across five different departments" needs formal purchasing and implementation rather than an individual's script. He compares current enterprise AI adoption to handing out PCs in 1983 or browsers in 1997, where distributing the tool did not transform operations.
↳ Follow the thread