Citizens expect digital government services to work the way the rest of their online lives do: one login, one experience, and no need to re-enter the same information across disconnected systems. For federal agencies, delivering that kind of unified experience is not just a technology challenge – it is a legal, funding, privacy, governance, and legacy IT challenge all at once.
For agencies and their industry partners, the challenge is modernizing services fast enough to meet citizen and employee expectations without weakening the oversight, data protections, and cybersecurity controls that government missions require, experts noted in a recent MeriTalk webinar.
Barriers go beyond technology
Truly unified services would mean “just a seamless, single experience across government,” said Kevin Walsh, director of information technology and cybersecurity at the Government Accountability Office (GAO).
Privacy laws, limits on data sharing, and appropriations tied to specific programs have historically limited that kind of integration, Walsh said. Agencies may have technical tools to connect systems, but they still must operate under legal frameworks that govern how data can be collected, shared, and used.
Legacy IT adds another layer of complexity. Federal systems – such as air traffic control – often predate the internet and were not built to integrate with modern platforms, Walsh noted.
Architecture and integration are key to unified services
Despite these challenges, it is possible to achieve more unified government experiences, the experts agreed. Architecture and design consistency are key, said Scott Ditch, lead solutions architect at OutSystems.
“Part of that unified government experience is that same look and feel across all those different systems and applications,” Ditch said. Enterprise architecture, he added, “guides you to that consistency, that security, that governance … [and] increases reusability.”
Jim Hutcherson, low-code/no-code chief technology officer at IBM, said agencies need to address three layers at the same time: the user interface, system integration, and the data and artificial intelligence (AI) layer.
“It’s a very complex problem that requires a lot of thought,” Hutcherson said.
Near-term steps can reduce modernization risk
Agencies can make progress without ripping out core systems all at once, Ditch said. They can start with user interface improvements, build consistency, and then move deeper into rules, data, and integration layers. Hutcherson said agencies should also prioritize access to application programming interface (API) layers, including building APIs where legacy systems lack them.
Procurement cycles and workforce gaps can slow that work. Traditional three- to five-year IT contracts often outlast the technologies they were designed to acquire, Hutcherson said, while agencies and contractors face pressure to keep technical skills current.
Shorter delivery cycles are critical, Walsh said, because long-running modernization programs risk delivering outdated capabilities before they reach users.
“By the time 10 years have passed, the technology is totally different,” Walsh said. “Focus on shorter sprints … more agile development. You just have to keep the overarching goal in view as you’re working on those shorter sprints.”
AI requires grounded data and guardrails
AI can help agencies move faster, but only when paired with strong data governance. Hutcherson said near-term AI value falls into two main areas: improving service delivery and increasing developer productivity.
Data quality remains central to AI performance, Walsh observed. “Garbage in, garbage out … still holds true, especially with AI,” he said.
Agencies need grounded systems that rely on verified information, Hutcherson said.
“If it’s not sourced correctly, don’t let it go out to … the employee or to the general public,” he advised.
Agencies also need governance structures that allow low-risk AI uses to move forward without weakening oversight. Hutcherson said that means setting clear risk tiers, pre-approving common implementation patterns, and building continuous monitoring into the design.
Guardrails must extend beyond the model itself. Agencies need to prevent sensitive information from entering AI tools, protect personally identifiable information, and ensure outputs are monitored, Ditch said.
Human involvement is critical, and agencies need to design it carefully. A human reviewer can become a rubber stamp if an AI system is right most of the time, Walsh said. To avoid this, agencies should structure review processes so employees must apply critical judgment, rather than simply approving an AI-generated answer.
Mission outcomes should drive measurement
Agencies should measure modernization by mission outcomes, not just system performance, Walsh said. Transaction speed and processing volume matter, but the better questions are whether veterans receive prescriptions faster, imports move more efficiently, or citizens complete services with less friction, he noted.
“I would really love to see mission-driven metrics and outcomes,” Walsh said.
View the full discussion.