Plato 完成 1600 万美元 A 轮融资，由 Nava Ventures 领投，Gradient Ventures、Transpose Platform 以及微软 AI 负责人 Mustafa Suleyman 跟投。该公司向专业人士付费，让他们将自己的工作重现为模拟环境，供 AI 实验室训练和测试智能体。Plato 累计融资达到 1900 万美元。
Plato raised a $16M Series A led by Nava Ventures, with Gradient Ventures and Microsoft AI chief Mustafa Suleyman joining. It pays professionals to recreate their jobs as simulations where AI labs train and test agents. Total funding is $19M.
这家总部位于旧金山的初创公司 Plato 专门构建虚拟工作场所，让 AI 智能体能够练习真实工作，公司于周三宣布完成 1600 万美元 A 轮融资。Nava Ventures 领投，Gradient Ventures、Transpose Platform 以及微软 AI 负责人 Mustafa Suleyman 参投。
Plato, a San Francisco startup that builds pretend workplaces where AI agents can practise real jobs, has raised a $16M Series A, the company said on Wednesday. Nava Ventures led the round, with Gradient Ventures, Transpose Platform and Microsoft AI chief Mustafa Suleyman joining.
Plato 复刻真实工作的软件和任务，涵盖客户支持、金融、会计、人力资源和医疗行政等领域。AI 实验室利用这些复刻环境，通过试错训练智能体，并测试其表现，而无需接触真实资金或真实记录。
Plato copies the software and tasks of real jobs, such as customer support, finance, accounting, HR and healthcare admin. AI labs use these copies to train agents by trial and error, and to test how well they work, without touching real money or real records.
“掌握真实复杂工作源于亲身执行这些任务的经验，”Plato 联合创始人兼首席执行官 Robert Farlow 表示。
“Mastering real complex work comes from experience doing those tasks,” said Robert Farlow, Plato’s co-founder and chief executive.
公司不仅考核最终结果，还检查智能体达成结果的过程，包括是否遵守规则。Plato 称，这能防止智能体通过钻测试空子取得高分，而非真正胜任工作。其投资人 Suleyman 曾主张，AI 在发布前应证明处于可控状态。
The firm checks both the end result and how an agent got there, including whether it followed the rules. That stops an agent from scoring well by gaming a test rather than doing the job properly, it says. Its backer Suleyman has argued that AI should be provably under control before it is released.
Plato 成立于 2025 年，累计融资现达 1900 万美元。新资金将用于支付专业人士重建其工作模拟环境，并派遣 AI 智能体执行真实任务以积累更多经验。
Plato was founded in 2025 and has now raised $19M in total. It will use the new money to pay professionals to rebuild their own jobs as simulations, and to send AI agents out on real tasks to gather more experience.
对这类训练场的需求正在增长。本月，AI 训练公司 micro1 宣布将投入 10 亿美元购买企业数据，构建类似的智能体练习环境。6 月，旧金山的 Patronus AI 融资 5000 万美元，专注于在模拟环境中对智能体进行压力测试。
Demand for such training grounds is growing. This month, AI training firm micro1 said it will spend $1B on company data to build similar practice environments for agents. In June, San Francisco’s Patronus AI raised $50M to stress-test agents in simulated settings.