Tools / Analysis
OpenAI’s LifeVal asks what AI can take off your hands
A useful assistant should leave you with less work. A new evaluation makes that ambition more concrete.
On 2 November, OpenAI presented LifeVal, an evaluation of everyday assistance. Its task selection combines time-use research with a survey of 3,995 people about where they want help. The point is practical: progress should show up in responsibilities handled, not only answers produced.
From time spent to help wanted
The starting points were the American Time Use Survey and Multinational Time Use Study. These catalogue activities people spend time on; the additional survey helped prioritize assistance. LifeVal includes household administration, education and shopping. OpenAI says it has used the evaluation internally over the past year.
In one published example, the assistant shortlists kindergartens and returns two preferences for the parent to decide. Completion still includes a human decision.
Sources: OpenAI [1] · U.S. Bureau of Labor Statistics [2] · Centre for Time Use Research [3]
A finished task needs a visible handoff
Our reading is that this changes the useful buying question. Instead of asking whether an assistant sounds clever, ask what remains for you to do after it stops. For example, a shopping shortlist and an order placed are different outcomes. Both can help, but they need different permissions and different evidence of completion.
That is the standard we would apply to an everyday agent: show the result, expose any assumptions and make the next decision easy to review. Keep irreversible actions under explicit control. A benchmark can help choose what to test; your own workflow reveals whether the assistant actually removes effort or merely moves it into checking its work.
Sources: OpenAI [1]
SOURCES & METHOD
The record behind the article
Checked against OpenAI’s announcement and the two linked time-use studies on 3 November 2026. No independent evaluation was run. The practical criteria in this article are our interpretation, not a description of LifeVal’s grading rubric.
- LifeVal OpenAI · 2026-11-02
- American Time Use Survey U.S. Bureau of Labor Statistics · 2026-11-03
- Multinational Time Use Study Centre for Time Use Research · 2026-11-03
