California Deploys Poppy AI Platform Across State Agencies in Record Time
What began as a statewide chatbot experiment has evolved into a versatile platform that can be managed and adapted to new technologies in real time. “What started as an effort to create a statewide chatbot quickly evolved into something exponentially more useful — a platform that could be managed and adapted to new technologies in real time,” said Shera Mui, deputy director of CDT’s Platform Services. Mui added that the team decided early that building large language models in-house was not the best path. “We discovered early on that we didn't want to create large language models; that was not going to be the smart thing to do,” she said.
To test Poppy, CDT granted early access to a cross‑section of 70 departments, each with around 100 users. The goal was to surface real‑world use cases and refine the system’s governance framework. The platform’s design lets users keep all data in a single secure location, eliminating the need to store information in multiple places. “We are able to offer multiple models from one location that keeps all of your data, all of your information secure, and you don't have to have it in four different places,” Mui explained.
The platform has already delivered measurable benefits. Dr. Lucy Andrews, a scientist with the California Department of Water Resources (DWR), said the AI tools have cut labor‑intensive work from weeks to an hour or 90 minutes. “We invest tens of millions of dollars every year in scientific work, but have lacked for a long time a way to track our scientific contributions,” Andrews said. She added that the new system allows DWR to trace public investment more accurately and demonstrate accountability.
While Poppy is powerful, CDT emphasizes that human oversight remains central. “No decision is left to an AI system and that humans maintain control of all outcomes,” the department said.
Early usage data from the statewide launch shows a strong appetite for document summarization, policy analysis, and complex code creation. Mui noted that some departments are interested in image‑generation features for future releases, but that additional work is needed to map a long‑term roadmap.
To support expansion, CDT requested $1 million in funding during the budget process. The money would help add new features, refine data sets, and foster cooperation among departments. “There’s a lot of instances where departments share data, and we have it all stored at different places, and so having Poppy be able to access some of those data spaces eliminates the need for duplication as well, and we can collectively put our resources together to really try to … clean up and be more cohesive as a state rather than at a department level,” Mui said.
The Poppy rollout aligns with California’s broader AI strategy, which includes recent legislation aimed at ensuring responsible, transparent, and secure AI adoption in public agencies. The state’s AI governance framework emphasizes privacy protection, civil liberties, and accountability.
As Poppy moves beyond the pilot phase, CDT plans to continue refining the platform’s model selection, data hygiene, and inter‑agency collaboration. The department also intends to monitor the environmental impact of running AI workloads, noting that DWR staff are mindful of water usage for data centers.
In summary, California has successfully deployed a multi‑model AI platform across state agencies, achieving significant efficiency gains while maintaining strict governance. The next steps involve expanding capabilities, securing additional funding, and ensuring that human oversight remains the final decision maker.