Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Technology Ai topic
No spam. Unsubscribe anytime.
Board hears generative AI primer and pilot ideas for contact centers; staff to bring governance policy
Summary
CalPERS received an overview of generative AI capabilities, deployment examples and risks from Accenture and internal staff; the board heard proposals for a pilot in the contact center and asked staff to return with governance and data‑security rules before broader rollout.
CalPERS staff and Accenture briefed the board on generative artificial intelligence (GenAI), its near‑term uses, risks and a proposed pilot to test contact‑center applications. Accenture’s public‑service lead explained common GenAI use cases—language translation, automated content generation, agent assistance and post‑call summarization—and demonstrated rapid prototypes, including multilingual content generation and an example of summarizing and translating a handwritten school assignment.
Staff described operational pilot options for the CalPERS contact center: using GenAI to (1) draft or summarize call notes (post‑call writeups) and (2) provide agents with suggested responses and knowledge‑article links in real time to improve first‑contact resolution and reduce handle time. Stephen Loverson and Accenture presenters demonstrated how enterprise models can be run behind firewalls so that data do not leave secured systems, and how systems can return source citations to reduce “hallucination” risk.
Speakers highlighted risks: biased outputs, inaccurate or “hallucinated” results, data‑leakage if public consumer models are used, regulatory and intellectual‑property questions, and the energy/compute cost of large models. Accenture emphasized responsible‑AI practices and recommended a governance framework before scaling: documented mandate/charter, access controls, model‑validation and testing, privacy and PII controls, training for staff and a staged pilot with monitoring.
Board members asked for a clear schedule and governance proposals. Several directors asked that staff present draft policies (use rules for enterprise models, prohibited uses, and handling of personal data) to the board or to the Finance & Administration committee before broad deployment. Staff said they will bring a governance policy and the contact‑center pilot design back to the board and begin limited, internally secured tests only after appropriate controls are in place.
