Derrick Pledger is the chief digital and information officer at Maximus, where he focuses on driving technology strategy, delivery operations and digital modernization activities for the organization.
For years, federal agencies have discussed artificial intelligence (AI) in terms of pilot programs and potential. But 2025 marks a turning point: we now see AI in the critical path of multiple missions that people depend on every day. The shift is clear: AI is no longer a standalone experiment; it is embedded within secure cloud environments, strengthened by reliable data, and paired with defined human decision-making roles – part of the operational infrastructure of government. Across the missions we support, meaningful progress occurs when AI is integrated into the systems that already carry the weight of public service delivery. The result is measurable improvement in how quickly, accurately, and consistently the public receives the services they rely on – not simply new technology in production.
Consider veterans’ disability benefits. For years, examiners and clinicians have navigated enormous volumes of medical evidence for each disability determination. A single claim can involve hundreds or thousands of pages. When demand spikes, backlogs grow quickly, and veterans experience that through longer waits and increased uncertainty about their claim.
Working with our Veterans Evaluation Services (VES) team, AWS, and government partners, we built an AI-enabled workflow that streamlines the process and improves the experience. Operating within a FedRAMP-authorized cloud environment built to manage sensitive veteran health records at scale, AI services now read and structure unstructured medical records, highlight the information examiners need most, and surface patterns linked to specific conditions and rating criteria. Human experts still make every decision; what changes is the quality of what they see when they sit down with a file. In the first year, the system processed more than 400 million documents – roughly 10 billion pages – and now routinely handles 8 to 10 million pages per day.
For veterans, the impact is direct: faster decisions, more consistent outcomes, and greater confidence that their case received a complete and timely review. Examiners spend less time searching through records and more time applying clinical judgment, enabling teams to absorb surges in demand without sacrificing quality.
In national security and cyber, the pressure is different, but the pattern is the same: deploy new capabilities quickly while hardening systems against sophisticated threats. Hackathons and rapid engineering sprints bring operators, engineers, and security teams together to solve problems in days instead of months. Much of what gets labeled “innovation” never makes it past security and accreditation. What matters now is the work that clears those hurdles and still moves quickly.
When we bring AI-enabled tools into that environment, we compress the time between an operational problem and a secure capability in production. Red-team simulations that used to take weeks now run in hours, and telemetry analysis surfaces anomalies before analysts do. For the public, that means stronger defense systems, faster mitigation of threats, and reduced risk to critical infrastructure.
In citizen-facing services, AI now supports routing, summarization, and decision support for agents in contact centers and digital front doors. A customer service agent who used to spend the first two minutes of a call pulling up case history now sees customer information automatically displayed before they say hello. Our work with AWS, Salesforce, and Genesys – grounded in Total Experience Management (TXM) frameworks – focuses on that level of specificity: shorter calls, more issues resolved on first contact, better satisfaction scores. People get clearer answers with less friction, even during peak demand.
As we move into 2026, the difficult part won’t be finding new AI use cases; it will be turning what worked in 2025 into a repeatable way of working. The choice is straightforward: standardize the playbook or slide back into one-off pilots that never reach scale.
For my teams, that means continuing to move toward shared platforms, stronger governance, and continued investment in the people who operate these systems. When we execute on that approach, AI becomes part of how government responds – not a one-off deployment. The result is lasting impact: shorter backlogs, faster decisions, and services that meet the urgency of the moment.