LobsterAI Report: Office Agent Task Size Up 3.2x in Six Months, 77% Dispatched via WeChat
huangyun_122 · x · 2026-10-19
LobsterAI released its monthly China Office Agent User Behavior Report (September edition) with notable data:
- Task scale: average tokens per task grew from 63K in March to 202K in September, a 3.2x increase — users now trust agents with long, multi-tool workflows.
- Dispatch channels: 77% of tasks are assigned via WeChat, 14.3% via Feishu, 6.5% via QQ; usage peaks 7-11 AM as commuters delegate work on the way to the office.
- Caching economics: DeepSeek V4 models hit 95%+ cache hit rates with cache priced at just 2-3.3%, so effective costs are only 4.6-7.5% of list price — versus 35.1% for GLM 5.3 Flash and 20.2% for Kimi K3, a stark cost gap.
- Use cases: programming/IT remains dominant, with design and content creation rising; the report also covers holiday-week usage shifts.
Related event: LobsterAI's first China office Agent report: tasks grow 3.2x in six months(8 posts)→
More from coding & agent
- A year running a multi-model router: 5 design choices that survived, 3 that died — AIGODSEND · 2026-10-19
- Stack Overflow for Agents launches in beta with 10k+ posts asked and answered by agents — wateriscoding · 2026-10-19
- DiceDB author proposes a standardized event protocol so AI agents can act proactively — arpit_bhayani · 2026-10-19
- Daybreak helps restore OpenCode users hit by an attack, refunding OpenAI API credits — fouadmatin · 2026-10-19
- Celesto Claims CI Runner With 96% Lower Per-Minute Cost Than GitHub Actions — aniketmaurya · 2026-10-19
- Reddit asks: which tasks can your agent do that you still do manually anyway? — Glittering812 · 2026-10-19