Department of Labor Chief Information Officer (CIO) Mangala Kuppa has a clear yardstick for artificial intelligence (AI): Show her the value.
Labor has spent the past several years building the governance, tools, workforce literacy, and processes for responsible AI use across the department. As demand grows, Kuppa is focused on what AI projects actually accomplish.
“We never looked at AI as a race against how many use cases we have,” Kuppa told MeriTalk. “It’s about how much value are we generating.”
One Labor implementation offers a concrete example. An AI use case within the Employee Benefits Security Administration (EBSA) is processing about 330,000 files and saving an estimated 21,000 hours, she said.
That focus on measurable outcomes also shapes how Kuppa sees the CIO role.
“I think we have to shift our focus from service levels to value generation as a measure,” she said. “For me, I see our role as a change agent.”
Start with what improves for the public
Asked to rank 10 competing priorities for fiscal year 2027 (FY27), Kuppa put digital government and citizen experience first. Cybersecurity and risk management came second, followed by AI and machine learning in third and workforce transformation in fourth.
“Any technology that we implement … has to result in some significant improvement in the service we provide,” Kuppa said.
She applies the same test to AI.
“AI must work for workers. AI must work for our public in the way that we implement AI,” she said.
Labor’s progress on AI over the past year may be the one thing that surprises people most about the department, Kuppa believes.
“Labor may not be the first department people think about when they’re thinking about AI, but the maturity of our governance process … and the implementation approach we took – and the progress we made – I think that would be surprising for people,” she said.
Reimagine the workflow with AI assisting humans
Kuppa said Labor organized its AI approach around nine pillars, including literacy, access to tools, pilots, continuous monitoring, and measurement.
Labor also uses a “use case ideation framework” that brings technologists together with the business units performing the work. Together, they map how a process operates today, explore where AI could assist, estimate the cost and return on investment, and give business leaders the information they need to decide whether to move ahead.
“To realize value, AI has to embed into workflows while ensuring human oversight of AI output,” Kuppa said.
Labor is pairing targeted implementations with secure AI tools employees can use directly. Kuppa said active users appear to be saving about two hours a day as they apply AI to routine work.
“If you just use that math and extrapolate the number of people … then you start seeing the real value,” she said.
When one subagency needed help transcribing large numbers of audio files, for example, Labor made an AI-enabled transcription capability available more broadly through its internal Job Enrichment Center.
“That self-service approach we’ve taken is tremendously valuable,” Kuppa said.
“We have made a lot of progress,” she said, “but I would still consider us as scratching the surface for AI.”
Address the talent gap
Labor already has demand for more AI work. Kuppa’s immediate concern is having enough skilled people to meet it.
“The talent gap … is truly the operational gap,” she said. “It’s not the ideas. It’s not the tools any longer.”
Labor initially concentrated AI expertise in a centralized team, alongside centralized governance. Kuppa now wants to bring dedicated expertise to support the department’s subagencies.
Her target is a few dedicated AI consultants per subagency. The goal is to give those specialists enough familiarity with each organization to understand its work, identify opportunities, and help employees rethink their workflows.
Labor is using traditional hiring alongside the Office of Personnel Management’s U.S. Tech Force program to build that capacity. Kuppa says retention will be just as important.
“All of these technologies need people with know-how,” she said. “It’s one thing to hire, but you also have to retain people because there’s so much demand out there.”
Build cyber defense for machine speed
AI is also changing how Kuppa thinks about cybersecurity.
Her first move has been to bring security expertise closer to the front end of technology decisions. Labor’s cybersecurity, technology, and operations leaders are now involved earlier as the department evaluates emerging technologies and designs AI architectures.
“Our cyber teams have to be elevated to front and center technology decisions,” she said.
In addition, agencies must reassess long-standing cyber practices against AI-driven threats, Kuppa says. AI can help attackers identify and combine vulnerabilities and exploit systems continuously at a scale that puts new pressure on cyber defenses, she noted.
“If you have AI agents trying to exploit your environment 24/7 in the cyber space, you have no option but to use AI to defend it,” Kuppa said. “A non-human entity working 24/7 coming at a scale to exploit your environment has to be defended with that same level of automation.”
“It’s time to reimagine some of the work we are doing in the cyber space,” she added.
Kuppa pointed to Authority to Operate (ATO) processes, continuous diagnostics, and patch management as areas worth examining. The goal, she stressed, is not to abandon practices that have protected agencies for years. It is to determine whether the effort is still aligned to the risk.
With patch management, for example, some vulnerabilities may warrant immediate action while others can be handled according to their actual risk. That kind of prioritization could give cyber teams more room to focus on the highest-risk threats.
Keep the engineering discipline
For all her emphasis on reimagining work, Kuppa also has a warning for agencies moving quickly with AI: Don’t discard the engineering practices that got federal IT this far.
Requirements still matter. Architecture reviews matter. Testing matters.
“It is important not to lose the discipline, the engineering discipline that we all built over the years in developing systems,” she said.
AI’s faster development cycles make that discipline more important, Kuppa said.
She wants teams to move quickly while still taking enough time to test whether an application works reliably and produces results worth sustaining.
“The modernization problem continues to be making sure your requirements are solid,” she said.
Business leaders must view IT as a strategic partner and bring IT into the conversation early on
If Kuppa could remove one source of friction heading into FY27, she would tackle technology acquisition.
“I think a lot of efficiencies can be gained in how we go about procuring technology,” she said.
Kuppa also wants business leaders to bring IT into the conversation earlier, while they are deciding how work should be done instead of after they have settled on a technology request.
“The realization that IT is not just a tool, it’s your strategic partner, has to happen in the business units,” she said.
That partnership mindset also helps explain why Kuppa has stayed in government far longer than she expected.
Kuppa came to Labor from the private sector 16 years ago thinking she might get bored and leave government within two years.
Instead, she stayed. She found strong technical teams, a steady stream of new problems to solve, and employees deeply committed to the department’s mission.
“We have a lot of talent in the federal government,” Kuppa said. “There is truly a dedicated workforce that goes above and beyond to actually fulfill the mission.”
For Kuppa, that workforce remains central to what comes next – whether Labor is scaling AI, reworking cyber defenses, or changing how IT and the business work together.