Federal agencies are rapidly expanding their use of artificial intelligence (AI), but SAP’s Jamison Braun said progress should be measured by mission results rather than the number of pilots underway.
Braun, senior vice president and managing director for SAP’s U.S. public services business, said the more consequential measures are whether AI removes friction, gives employees time back, and improves a service or mission outcome.
“AI amplifies readiness; it doesn’t create it,” Braun said in a recent interview with MeriTalk. “The next phase belongs to agencies that end up turning trusted insights into governed action at mission scale.”
The Office of Management and Budget’s 2025 AI use case inventory illustrates the scale of that ambition, Braun explained.
The inventory lists 3,611 individually reported AI use cases across all stages of development. It includes 56 total agency submissions, including 41 submissions reporting individual AI use cases. However, Braun cautioned against treating the volume as proof of operational success.
“That’s activity,” he said. Although the increase demonstrates momentum, Braun added, it is “not proof that there was adoption or that the outcome scaled. We haven’t been able to see the outputs that show the value.”
At the same time, Braun cited a study underwritten by SAP in which only 23% of federal chief information officers (CIOs) and executives described their business systems as advanced or fully modern.
“AI ambition today is actually moving faster than the operating foundation,” he said.
Start with the outcome, not the model
For Braun, the strongest path from experimentation to production begins with a specific operational problem. He said agencies should identify a high-volume workflow where a backlog, reconciliation burden, supply exception, auditability problem, or service delay is visible and measurable.
“Start small and expand,” Braun said. “You’ve got to deliver one measurable outcome, and then you have to scale what works.”
That approach shifts the definition of progress from AI activity to mission value. A model may generate a useful answer, Braun explained, but operational AI must also understand the live business context, apply existing rules and permissions, and help a person take the next step.
“The near-term opportunity, and in fact the long term, is not removing people from high-consequence decisions,” he said. “It’s detecting that friction earlier. It’s reducing the repetitive work, and it’s giving public servants more time for judgment and service.”
What production looks like
Braun pointed to Queensland, Australia’s Department of Transport and Main Roads as an example of AI embedded in operational decision-making.
“Queensland’s Department of Transport and Main Roads uses an AI-powered SAP platform right now to evaluate millions of pavement investment scenarios,” he said.
The department’s cloud-based Asset Deterioration, Valuation, and Investment Calculation Engine (ADVICE) uses SAP AI Core to model treatment strategies and compare investment options across Queensland’s road network.
Braun also described sensors on transport and road vehicles automatically sending notifications about potential problems, which he said reduced response and repair times from 11 hours to four.
The broader benefit, he said, is allowing “citizen governments to report faster, be more precise, and have auditable investment decisions.”
Escaping pilot purgatory
In Braun’s view, three gaps frequently keep AI initiatives in what he called “pilot purgatory.”
The first is missing mission context: The technology may not understand a workflow’s approvals or exceptions. The second is missing the connection: Fragmented systems prevent it from seeing the full operational picture. The third is missing governance: Leaders cannot trace an output or control the action a system may take.
Braun argued that SAP’s enterprise systems can help address those gaps.
“SAP is uniquely built to work at the point where government decisions become transaction services and mission action – all grounded in the authoritative dataset,” Braun said.
Braun also highlighted SAP’s OneGov agreement with the General Services Administration (GSA).
GSA announced the deal in December 2025, giving agencies discounted access to the company’s database, integration, analytics, and cloud solutions for federal agencies.
GSA negotiated a deal under which SAP is offering discounted pricing, and GSA estimates those discounts will save the government $165 million.
Advice for mission leaders
For government CIOs or mission leaders who are trying to move quickly but responsibly with AI, Braun said they must “move faster, faster.”
“You have to move incredibly quickly by narrowing the problem, not by lowering the standard,” he said. “Find that one source of mission friction, establish the baseline and guardrails, put an accountable leader behind it, and build for production.”
He recommended mapping the decision, data, handoffs, exceptions, and authority involved in a workflow before automating it. Agencies should establish risk-based boundaries for AI actions, maintain human authority where consequences are high, and make outputs traceable and defensible.
For a practical first step, Braun advised bringing the mission owner, process owner, data owner, security lead, and workforce lead together around one bottleneck, one point of friction, and one measurable outcome. Then, he said, that cross-functional group can “solve it in 24 hours.”
The ultimate goal, according to Braun, is to leverage AI to turn insight into action.
“We have enough AI that reports the news, that tells us what already happened,” he said. “We need AI tied to those mission outcomes. That helps us create white space to make citizen government more effective.”
“The goal is not autonomous government,” Braun added. “The goal is government that can see clearly, act responsibly, and deliver when people need it, with AI doing more of the invisible work, so that public servants can do more of the human work.”