October 20, 2026
Small team, big impact
How the California Workforce Development Board is putting POPPY to work
SACRAMENTO, Calif. – At the California Workforce Development Board (CWDB), a small team with a wide-ranging mission has found a new way to make its resources go further.
With just over 80 positions, the California Workforce Development Board is accustomed to doing a lot of work with limited resources. POPPY, the state’s generative AI digital assistant, is helping the department expand that capacity by turning time-consuming tasks into work that can sometimes be completed in minutes. It’s also helping staff rethink processes that once took months.
“We’re heavily constrained by limited staffing, and we have a lot of mandates,” said Emily Sunahara, CWDB deputy director of operations and policy. “POPPY has helped lighten that workload.”
CWDB participated in the year-long POPPY pilot and has been exploring ways to build tools around the expertise its employees already have. Those uses range from research and workforce development to reviewing duty statements, supporting professional development, and improving internal processes.
For Deputy Counsel Ben McDonald, POPPY has become particularly useful for one of CWDB’s most important responsibilities: supporting its 42-member board.
The board includes appointees from across California. As it is governed by the Bagley-Keene Open Meeting Act, board meetings can generate hours of discussion, presentations, and other information that must be captured and shared. Producing useful summaries from all that material requires significant staff time.
McDonald developed a POPPY tool with detailed instructions for doing much of that initial work. He can provide the meeting agenda, transcripts and presentations and have POPPY produce a summary that captures the essential information.
A task that could consume eight hours or more can now produce an initial result in about a minute.
He said that time savings doesn’t eliminate human review, it makes more time for it.
“Accuracy is very important,” McDonald said. “You make sure it is what you want to say and nothing more.”
Overall, he estimates POPPY can save about 75 percent of the time required for some of his work. For a legal office with only two people, those hours matter.
“It lets me spend time on things that deserve my time,” McDonald said.
From six months to six weeks
CWDB has also found that POPPY can help improve processes that don’t directly use generative AI.
During the POPPY pilot, the department built a specialized model to help staff design a new grants process. Rather than using AI to automate an existing task, CWDB used POPPY as a subject-matter resource while staff developed a new process designed to remain compliant with state laws, regulations and policies.
The work required CWDB to invest time upfront, building a model with comprehensive subject-matter knowledge and refining the prompts used to guide its responses.
The department launched the redesigned grants process in early 2026. Grant package processing times dropped from about six months to approximately six weeks.
For Sunahara, the experience demonstrates a broader potential because POPPY helped staff design a more efficient process.
It also demonstrates how taking the time to build a specialized model can produce benefits beyond completing a single task. In this case, POPPY helped CWDB rethink the system behind the work.
Building AI around
One of POPPY’s strengths for CWDB is the ability to create tools that work within a defined body of information.
For McDonald, knowing what an AI tool has access to and what it doesn’t is critical to trusting the results. The technology is most useful, he said, when it is paired with strong subject-matter knowledge and clear instructions.
“POPPY means nothing if it isn’t reliable,” he said.
That approach has also turned POPPY into a learning tool. Staff can use it to explore a subject, identify areas where they need more knowledge, and even improve the way they interact with AI. Sunahara said the model has helped staff recognize what they could add to their prompts to produce better responses.
Of course, this results in faster work, but it is also an opportunity to develop both the technology and the people using it.
For example, CWDB has explored using POPPY to crosswalk duty statements and professional goals to training courses in support professional development. McDonald has also used it to examine governance questions, such as building a board seat calculator based on formulas for board membership outlined in statute.
That versatility is especially valuable to a smaller organization.
“All of us have some repetitive tasks that could be automated,” Sunahara said. “POPPY can help streamline those workloads. Especially for small entities with limited resources, this was an affordable option that has not only alleviated some of the strain but contributed to improving the way we work.”
The technology
CWDB expected technical ability to be one of the biggest challenges of introducing generative AI. Instead, Sunahara said, the technical implementation went smoothly.
“I thought technical skills training required for a new IT product would be the biggest hurdle, but it was people’s belief about generative AI and their willingness to us it,” she said. “Changing hearts and minds was much more significant.”
CWDB had executive support for the pilot as well as champions and subject-matter experts willing to experiment with the technology for non-IT use cases. Looking back, Sunahara said she would have put even more emphasis on change management to address fears earlier. Providing exposure to use cases and training has supported increased buy in. Implementing structured peer-to-peer learning is also helping shift beliefs while building institutional knowledge.
McDonald sees that education as essential. Government organizations need people who understand both what generative AI can do and where its limitations lie.
“You have to prepare people to use it and know what it can and can’t do,” he said. “You can’t prepare for the impact without deeply understanding what it can do.”
Sunahara also sees AI skills becoming increasingly important to employees themselves.
“We wanted to be bold,” she said. “People may lose opportunities to advance their careers if they don’t learn to use AI.”
That means successful adoption requires more than giving employees access to new technology. It means showing them how the technology can make their work better and developing employees who can become ambassadors for responsible AI use.
A partnership with CDT
Both Sunahara and McDonald credit the California Department of Technology (CDT) team with making CWDB’s adoption easier.
“Fantastic,” Sunahara said of the experience. “They made themselves available to us above and beyond expectations.”
McDonald described the relationship as a two-way street. When CWDB encountered a bug, the team reported it to CDT, and it was corrected quickly.
That collaboration is something CWDB would like to expand beyond its own organization. Sunahara and McDonald see opportunities for departments using POPPY to engage with one another, share best practices, and troubleshoot common challenges.
Their experience also offers a straightforward lesson for other state and local government organizations considering generative AI: start with the work employees are already doing.
That has meant reducing grant package processing times, transforming hours of meeting preparation into minutes of initial work, and giving employees more time to focus on responsibilities that require their attention.
The results show what can happen when government entities look beyond using generative AI to complete individual tasks and begin exploring how it can improve the systems and processes behind those tasks.
For CWDB, POPPY isn’t just helping a small team work faster. It’s helping that team find better and smarter ways to work.