AI agents cut task time 87% and reshape knowledge work scope
A study using Perplexity's Search and Computer products finds that autonomous AI agents perform 26 minutes of work per session versus 33 seconds for conversational search, reducing task completion time by 87% and cost by 94% compared to humans using Search alone.
Score breakdown
Benchmark your agentic tooling against these metrics — 87% time reduction and 55% lower dissatisfaction — as the paper establishes a concrete empirical baseline for what autonomous end-to-end execution delivers over conversational search in real production settings.
- 01Computer performs 26 minutes of autonomous work per user session; Search performs 33 seconds.
- 02On matched tasks, Computer reduces completion time from 269 to 36 minutes.
- 03Estimated time and cost are lowered by 87% and 94%, respectively, versus humans using Search alone.
A paper by Jeremy Yang, Kate Zyskowski, and Noah Yonack studies the transition from conversational AI assistants to autonomous agents using production data from Perplexity's Search and Computer products. The researchers use sessions with near-identical initial query pairs as natural experiments — treating them as the same underlying task attempted with both products — to isolate the effect of autonomy. Computer performs 26 minutes of autonomous work per user session compared to 33 seconds for Search, automating task decomposition and execution that Search users would otherwise orchestrate manually. This autonomy also shifts the distribution of follow-up queries toward higher-order activities such as verification and extension, and reduces per-query dissatisfaction rates by 55% relative to Search.
On matched tasks, Computer reduces completion time from 269 to 36 minutes, lowering estimated time and cost by 87% and 94%, respectively, compared to humans equipped with Search alone.
On matched tasks, Computer reduces completion time from 269 to 36 minutes, lowering estimated time and cost by 87% and 94%, respectively, compared to humans equipped with Search alone. Beyond efficiency, the study finds that Computer changes the scope of work users attempt: queries more frequently cross occupational boundaries, require higher-order cognition, draw on broader expertise, and take the form of composite tasks that bundle interdependent subtasks into a single query. The paper also notes that Computer unlocks work activities essentially absent from Search usage among the same users, suggesting that agent autonomy expands not just the speed but the frontier of what knowledge workers attempt.
Key facts
- 01Computer performs 26 minutes of autonomous work per user session; Search performs 33 seconds.
- 02On matched tasks, Computer reduces completion time from 269 to 36 minutes.
- 03Estimated time and cost are lowered by 87% and 94%, respectively, versus humans using Search alone.
- 04Per-query dissatisfaction rates are 55% lower on Computer than on Search.
- 05Computer shifts follow-up query distribution toward higher-order work such as verification and extension.
- 06Computer queries more often cross occupational boundaries and bundle interdependent subtasks into composite queries.
- 07The study uses near-identical initial query pairs as natural experiments to compare the two products.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 8, 2026 · 15:36 UTC. How this works →