GSA Trains 3,000 Staff on AI as Agencies Rewrite Modernization Playbooks

GSA trained 3,000 employees across 60 AI sessions and measured a 30% increase in AI prompts and conversations. That is the concrete return the agency reports from its shift to what it calls continuous change. Whether a 30% prompt increase translates into better citizen services remains unmeasured in the agency's public account.

The numbers came from Lesley Briante, associate chief information officer for digital services at GSA, speaking on the panel "Scaling Smarter Government Services Across Agencies." "Technology is moving so quickly that I think the expectation now is that we have continuous change," Briante said. "AI is acting as an accelerator to that change."

What is GSA actually doing?

GSA is close to finalizing a technology roadmap covering the next three to five years. The plan focuses on digital transformation that mirrors how citizens use technology in their personal lives.

Concrete elements include:

• Automated help desk services and expanded self-service
• Shared platforms such as USAi and Login.gov for a seamless citizen experience
• A balanced-scorecard metrics framework covering strategic, operational and day-to-day measures
• A dashboard tracking AI tool usage and flagging where employee training falls short

The operational data works at a granular level. Weekly help desk analysis by geography recently identified that newly installed printers were generating calls because staff did not know how to map to them. GSA targeted training and communications at that single gap.

How is Maryland attacking technical debt?

Maryland's biggest problem is older than most of its workforce: mainframes. Melissa Leaman, deputy secretary and chief operating officer at the Maryland Department of Information Technology, said the state's five largest agencies — public safety, health and human services among them — carry what she called an "insane amount of legacy technical debt" on mainframes.

The state published a three-year IT master plan last summer. A separate, legislatively mandated state modernization plan forced a review of thousands of applications and systems to quantify that debt. The findings now steer priorities for the state's IT investment fund. Leaman noted the master plan "can tend to get outdated rather quickly" given emerging technology — a candid admission for a planning document.

Maryland's three other stated priorities: stronger cybersecurity across agencies, improved accessibility and usability for citizens, and sustainable IT operations.

What changed in procurement?

Maryland has rebuilt how it buys technology. Agencies must now run discovery with user-centered and user-experience feedback before purchase. Requirements packages are built from that feedback up front, not negotiated after award.

The state legislature passed legislation earlier this year permitting proof-of-concept procurements. Agencies can pilot emerging technology under one solicitation and move directly to award without a second competition. Maryland is preparing its first pilot under the new authority. Efficiency is the promise; no results exist yet.

How are vendors responding?

Kristin Seaver, vice president of government operations and strategy at GDIT, said her company moved from past-performance pitches to building demonstrable capability before proposals. "We need to create the demonstrable capability for that agency and their desired outcome before we even get to a proposal," Seaver said. The approach uses micro-solutions, proofs of concept, and lessons transferred from other agency customers in software development and AI.

Jerome Johnson, director of professional services for U.S. federal, defense and aerospace at AWS, said agencies now hit cost, schedule and performance targets while also satisfying front-end customer needs. On AI, he stressed model optionality — cloud or open source. AWS is pushing toward firm-fixed-milestone-price contracts on the professional services side, with requirements handled up front.

The through-line across GSA, Maryland, GDIT and AWS: outcomes first, continuous iteration second, and no assumption that any roadmap survives contact with the next technology wave.