Federal Agencies Shift AI Efforts Toward Mission-Specific Applications

Federal agencies are working to turn artificial intelligence from a collection of isolated experiments into mission-ready capabilities. At FedInsider’s Innovation Summit 2026 on Oct. 21 in Washington, D.C., officials from the National Institutes of Health (NIH), the U.S. Department of State, and the Library of Congress described a common challenge: agencies must identify operational problems first, then determine where AI can help. That approach moves procurement away from generic chatbot demonstrations and toward solutions tied to defined workflows, governed data, and outcomes that can be evaluated.

For contractors, this evolution changes how AI capabilities should be framed and supported. NIH Associate Director of Informatics Joe Croghan said agencies should not ask employees simply how they might use AI, explaining, “I think one of the responsibilities we have is to not ask people, ‘How do you think you could use AI?’ I don’t think they know.” Instead, technical teams need to identify where workflows break down, what data employees manage, and which tools can address those problems. Croghan said NIH has “pivoted away from the personal productivity that a lot of companies are providing” and focused on “purpose-built applications.” Vendors competing for federal work should therefore connect proposed capabilities to agency-defined needs and demonstrate how performance will be assessed.

Scaling these applications also depends on foundations that extend beyond software. NIH is working toward shared infrastructure incorporating policy, data governance, and financial oversight, with the aim of supporting promising projects without creating disconnected tools that are difficult to manage or scale. Dell Technologies sales engineer Cole Feely warned that federal data can be “extremely chaotic,” cautioning that agencies cannot put disorganized data into a powerful system and expect it to operate properly. He highlighted the need to organize relevant datasets and establish security, monitoring, and access controls early—especially when AI agents may retrieve information or perform tasks on employees’ behalf. This creates demand for integration, data preparation, security engineering, and operational oversight alongside AI products.

Workforce readiness and procurement flexibility are equally important to adoption. The State Department supports more than 70,000 employees across more than 200 overseas missions, making continuous training and feedback channels central to effective deployment. Training cannot be a one-time event: agencies need to update guidance as tools change and give employees ways to influence application development and rollout. The Library of Congress’s AI sandbox and preference for flexible procurement arrangements further signal interest in testing capabilities before broader commitment and adapting solutions as requirements evolve. Contractors can strengthen their position by proposing phased implementation, measurable evaluation, sustained training, and adaptable arrangements rather than fixed offerings that assume agency needs will remain static.

• NIH is emphasizing purpose-built AI applications over general personal-productivity tools, starting with workflow problems and agency needs.
• NIH’s planned shared infrastructure incorporates policy, data governance, and financial oversight to help promising projects scale without creating disconnected tools.
• Joe Croghan said agencies should identify workflow breakdowns and data needs rather than ask employees to invent AI use cases themselves.
• Dell Technologies representative Cole Feely warned that chaotic data can undermine AI performance and called for early dataset organization, security, monitoring, and access controls.
• AI agents that retrieve information or act on employees’ behalf increase the importance of permissions, monitoring, and controlled access.
• The State Department supports more than 70,000 employees across more than 200 overseas missions, creating significant workforce training and enablement requirements.
• State Department training needs to be continuous, with updated guidance and employee feedback as AI tools evolve.
• The Library of Congress AI sandbox and preference for flexible procurement point to opportunities for adaptable solutions and staged testing before broader deployment.
• Contractors should tie AI proposals to defined workflows and measurable operational outcomes, not rely solely on generic pilot demonstrations.
• Pairing AI tools with data integration, implementation, security, training, and ongoing support can address agencies’ broader adoption requirements.