If an agent finishes in 30s but review takes 5min, what did you actually automate?
djgoel123 · reddit · 2026-10-29
A developer deploying AI agents poses a core question: an agent completes a task in 30 seconds, but an employee spends five minutes verifying the result. The task is technically automated, yet the work hasn't disappeared — it moved into verification.
The author argues human review is still essential where mistakes are expensive; the real issue is designing workflows with the right level of oversight — which actions run automatically, which need approval, and which shouldn't be delegated at all. He asks peers: how do you measure actual time saved after accounting for review, corrections, and exceptions, and how do you decide what's safe to automate without approval?
Related event: AI agents shift work to verification instead of saving it(2 posts)→
More from AGI Musings
- GPT-6.1 computer use costs $0.16 per human-minute task, an effective $9.60/hr — bytebot · 2026-10-29
- When agents outnumber humans, their preferences will steer the economy — lfschiavo · 2026-10-29
- Millions of tweets, zero scenarios Shulman and Bostrom didn't already think through — abhiadesai · 2026-10-29
- AI Software Factory vs Dark Factory: two stages of autonomous software delivery — Pavan_Belagatti · 2026-10-29
- Is Google's claimed use of RSI the start of the singularity under Vinge's and Good's definitions? — Extraaltodeus · 2026-10-29
- 'AI is just a tool' misses the point: writing and print reshaped humans too — YogeshMalik · 2026-10-29