Baidu's Shen Dou: RMB 1,000/Month per Employee to Try LLMs, No Hard Targets — Nudge With Perks, Don't Mandate AI
Baidu's Shen Dou says mandating AI at work rarely works; Baidu instead gives each employee RMB 1,000/month to freely try any LLM, unbound to KPIs. He expects agents to land at scale in H2 2026, and Robin Li proposes "Daily Active Agents (DAA)" as a new metric.
Nudge with perks, no hard rules
Per IThome citing Huxiu (July 17), Baidu EVP and AI Cloud head Shen Dou said mandating AI at work rarely delivers; Baidu leads with positive guidance and perks. Its standing employee benefit remains in force — RMB 1,000 per month for each employee to freely try mainstream LLM products, any use, unbound to performance reviews — purely to lower the trial barrier. The logic: once employees feel AI's productivity value firsthand, regular use becomes a habit.
The call: agents landing at scale soon
Shen predicts general and vertical agents will scale much faster, with clear batch-adoption cases in H2 2026; 90% of work will involve deep AI participation and assistance rather than full replacement. He sees the next three years as a key window: vendors keep realizing the value of "full-stack," and rivals are building full-stack too — both an opportunity and greater competition for Baidu AI Cloud.
A metric shift: from tokens to DAA
On the numbers, Baidu's Q1 2026 AI revenue rose 49% year over year to 52% of revenue — the first time AI exceeds half. At Create 2026, Robin Li proposed "Daily Active Agents (DAA)" as a new metric, mirroring mobile's DAU. He argues token consumption — today's nearest-to-consensus metric — measures only "input/cost," not "output/return," while DAA focuses on how many agents are doing work and delivering results, closer to real value. Baidu pairs "perk-led adoption" with "DAA measurement" as a combined playbook for spreading and gauging AI.