Data entry is a really interesting one to me. We’ve been replacing a legacy (read > 1 MLoC with no tests), system on and off for a few years. The original system had a ton of double entry or manual data entry and the human error rate is noticeable. If GPT could have automated this with a similar or reduced error rate then we would have considered it a win.
The real win long term remains killing off manual entry any time it’s possible, but GPT offers a nice patch.
The real win long term remains killing off manual entry any time it’s possible, but GPT offers a nice patch.