Bayard Cowan, chief technology officer at the Department of Consumer Affairs (DCA), and the department’s web architect presented a proof‑of‑concept chatbot for the board’s website that uses AWS Bedrock and GenAI components. The demonstration emphasized data scoping, human review and compliance with state GenAI policy.

Cowan said the state has adopted an executive‑level and technical framework for GenAI use and procurement, and DCA has developed internal acceptable‑use policies and staff training. “Executive order number N‑12‑23 from the governor…,” he said, and added the California Department of Technology issued a technology letter (25‑01) in 2025 that, together with state laws, guides GenAI procurement and oversight. Cowan said DCA has completed staff training and an IT acceptable‑use policy and will perform a formal GenAI risk assessment and submit it to the Department of Technology before a public rollout.

The technical demo showed a chatbot limited to the board’s website content with search and matching that tolerates acronyms and informal phrasing. The web architect said “the responses are restricted to your data and your data only,” and demonstrated that every interaction is logged for later analysis and that staff can add alternate phrasings to the question bank. The presenters emphasized the tool is “highly tunable” but requires ongoing back‑office maintenance: content updates, analytics review and periodic training of canned responses.

Board members asked whether the system can prompt users for follow‑up information (a tree‑like flow); DCA said yes, the implementation supports guided follow‑up questions and branching paths. The presenters estimated a near‑term proof‑of‑concept roll‑out in about 90 days to a development server, with additional review of Google/Microsoft offerings and the Statewide Digital Assistant under development by the Department of Technology.

Board staff said a recently approved reclassification for a data analyst position will include ongoing monitoring duties for the chatbot. DCA reiterated that human oversight—sampling and review of chat histories and administrative review of answers—is required by state policy before the tool answers regulatory questions to the public.