In Brief

Putting AI to Work in Government

Written by Fed Gov Today | Nov 16, 2026, 5:40:36 PM


Presented by IBM & Carahsoft


Federal agencies have put generative AI into employees’ hands. The next challenge is making it part of how government delivers results. In a recent episode of In Brief, Jason Prow, Senior Partner, Service Line Leader, Data & Tech Transformation, Federal Market, IBM, joins Fed Gov Today’s Francis Rose to discuss how agencies can connect AI to everyday workflows—and prepare their data, infrastructure, and workforce to support it.

Moving AI Into Agency Workflows

Generative AI’s accessibility has helped agencies explore new ways to improve productivity and give employees faster access to information. But Prow says that focus can also distract from a larger opportunity: integrating AI into front-office and back-office workflows.

That shift begins with understanding the process an agency wants to improve. Before applying new technology, leaders should examine which steps are necessary, where work slows down, and what AI or automation could handle differently.

Prow describes an approach that goes beyond conventional process reengineering to consider whether individual steps can be eliminated. The objective, however, is better performance rather than simply fewer steps. Some processes become more effective by removing activities; others benefit from adding steps while redesigning how the overall work gets done.

Connecting AI to Reliable Data

For agencies operating a mix of legacy and modernized systems, integrating AI can appear particularly challenging. Prow points to advances in automation, orchestration, and integration that help organizations access data within existing systems.

Data virtualization offers another way to make information available for AI use. At the same time, improvements in data management and techniques that help establish meaning and relationships across information can strengthen the quality of the results.

Prow identifies the volume and variety of federal data as a continuing challenge. Access alone is insufficient: agencies also need confidence that the information supporting AI is suitable for the task.

Human review remains part of that equation. Prow describes using preprocessing techniques to direct information requiring analysis to a person, reducing the volume of material that reviewers must examine. As organizations evaluate output quality, they can make more informed decisions about where human attention remains necessary.

Planning for Infrastructure and Cost

AI adoption also introduces infrastructure and consumption decisions that agencies need to address early. Prow discusses balancing cloud resources with targeted local processing, depending on the workload.

For intensive AI inference, he says on-premises processing can be more cost-effective in many cases. That does not necessarily mean reversing an agency’s cloud strategy. A smaller physical footprint could support particular processing needs while other workloads remain in the cloud.

Managing costs requires several approaches. Prow points to monitoring token usage, selecting an appropriate model for a given question, and considering how models are managed and deployed. Together, those decisions can help agencies better understand and control the resources their AI applications consume.

Preparing the Organization to Scale

Technology decisions are only part of the preparation. Agencies also need mechanisms to govern AI, monitor its performance, understand the sources of its data, and evaluate the quality of its output.

“But more importantly, it requires a different operating model and a different skill set.”

— Jason Prow, IBM

Prow encourages agencies to begin workforce training and organizational planning now. Leaders should assess where AI could support mission-critical processes, what oversight those applications require, and how responsibilities may need to change.

The goal is to avoid reaching broader deployment several years from now only to begin discussing the people and operating structures needed to support it. Putting AI to work in government requires sustained attention to the process, the data, the cost, and the organization responsible for delivering results.

Three Key Takeaways

  • Start with the workflow. Examine the business process and identify opportunities to redesign or eliminate steps before applying AI.
  • Build confidence in the data and output. Better access, stronger data quality, and focused human review support practical implementation.
  • Prepare for scale now. Infrastructure planning, cost management, governance, and workforce training should develop alongside AI adoption.

📺 Watch the full interview