Provision Labs targets the data and evaluation layer for physical AI
The AMW Read
The report repeats the previously covered launch of a robotics data infrastructure entrant, without disclosed funding, commercial customers or a confirmed Pronto data-sharing arrangement.
Provision Labs targets the data and evaluation layer for physical AI
Pronto founder Anjali Sardana has launched Provision Labs in San Francisco to build data, evaluation and deployment systems for physical AI. The startup names Lachy Groom, General Catalyst, Bain Capital Ventures and Glade Brook Capital as backers. It has not disclosed its funding amount or announced its first commercial customers. The report says Sardana will lead Provision Labs while continuing as Pronto's chief executive.
Provision Labs enters the data infrastructure layer serving robotics developers. Its stated role is to help developers train and test systems that manipulate objects and complete physical tasks. That places the company between real-world demonstrations and the models that control machines. The market question is whether those demonstrations provide useful training material and credible tests under changing conditions. Investor backing establishes support for the venture; the report provides no customer evidence that its tools improve robot performance.
For builders and investors, data access and permission are concrete diligence questions. Pronto's optional Pronto Verified program records household service sessions, and the company says recordings are anonymized before AI training. Neither company has publicly confirmed that those recordings will flow to Provision Labs. Buyers should therefore verify the startup's data sources, permitted uses and evaluation methods before assuming it has access to household demonstrations. Worker consent, household privacy and retention rules also matter when assessing whether recordings can support commercial training.