The Florida Department of Management Services is implementing AI, specifically generative AI developed by KPMG, to modernize its job classification system and enhance operational efficiency across state agencies.
The Florida Department of Management Services is harnessing artificial intelligence to overhaul its job classification system, leveraging tools developed by global audit and consulting firm KPMG to enhance operational efficiencies across state agencies.
In an interview with The Capitolist, Arthur Higbee, a KPMG AI Specialist, detailed how the state has applied generative AI to modernize a 20-year-old job classification framework. The initiative is part of a broader trend where AI technologies are increasingly integrated into day-to-day government operations to streamline processes and improve service delivery.
Higbee explained how generative AI, a technology capable of generating text and data interpretations, is used to extract key elements from numerous job descriptions and fit them into a new, modernized framework. This AI-driven approach allows for a more efficient handling of vast data sets, which would traditionally require extensive manual labor. He further elucidated that the AI tools are adept at extracting data from extensive volumes of unstructured documents, such as PDFs, thereby reducing the manual effort typically involved in such an undertaking.
“The work that we’re doing with the Department of Management Services is really a perfect use case for generative AI and sort of the latest of what the technology can do,” Higbee said. “You can feed it PDFs or other types of documents, it can pull out information that you’re looking for.”
The technological aid comes with a review process involving multiple layers of human oversight to ensure accuracy and reliability. Higbee stressed the importance of human oversight in AI implementations, stating that though AI can perform initial data processing and categorization tasks, KPMG ensures that the results undergo several layers of human review to guarantee accuracy and appropriateness.
“It’s critically important to keep the human in the loop at all times,” Higbee urged.
However, the adoption of AI is not without its challenges. Higbee acknowledged that some politicians are apprehensive about new technologies, often due to a lack of understanding. To mitigate these concerns, KPMG has instituted a set of “trusted AI principles” that ensure ethical usage and data security, reflecting the firm’s commitment as a “responsible steward of data.”
“One of the first things we did was come out with our trusted AI principles, which outlines the 10 things that guide our use of AI,” he continued. “We’re a 150 year old firm. we audit tons of companies and government agencies and we’re responsible stewards of data and confidential data. We wanted to make sure that it’s locked tight. It’s completely secure.”
The conversation also touched on broader applications of AI within the state government, including areas like fraud detection and timesheet processing within the management of grants, in large reducing the administrative load on agencies, allowing staff to focus on higher-value tasks.
“We’re seeing this race towards adoption of AI and I think regardless of where the agencies are at, across the board, you’re gonna see that happen more and more,” said Higbee.
In January, The Capitolist spoke with Anthony Monaco, a Government Partner with KPMG who provided insight into AI’s function within agency work.
A separate KPMG AI model makes the management of grants more efficient, according to Monaco, aiding in tasks such as timesheet processing and fraud detection, which could lead to time and cost savings for the government. Moreover, the platform manages high-volume grant activities by transforming unstructured data into a structured format and applying intelligent automation for accurate information processing and fraud detection.
“We’re doing simple routines, and it doesn’t need to be very complex,” said Monaco. “An example that we use is that we found professionals charging for 25 hours in a day or more, to seek reimbursement. So as simple as that is, we’ve identified fraud in those situations.”