The bottleneck in automation was data, not the model
A record in which someone who pushed work automation forward on their own for a year, with no outside help, lays the process out in three stages. The first stage is collecting: piling five streams — email, calls, meetings, document files, and in-house system data — into one place. Here a person spends several hours only on setting the sorting criteria and leaves the sorting itself to AI, and before feeding in internal material they first check the security policy and the masking of personal information. The second is talking: instead of writing the manual out as a document, you narrate your working screen out loud while recording it and hand that over as is, and the AI writes the task guide for you, screen images included. The third is delegating: how far to hand over is called the degree of delegation, and AI is given only up to the draft while a person does the final publishing. Company-reported, AI writes about 80% of the draft and a person checks and decides on only the remaining 20%.
Key points
- Five streams — email, calls, meetings, files, in-house systems — collected in one place
- A person sets the sorting criteria; the sorting itself is left to AI
- Hand over a recording of the screen and it writes the task guide for you
- Only the draft is handed over; a person does the final publishing
This is a note: my own write-up after reading several sources. It has no source link; events checked against a source are kept as records.