Challenge

Federal agencies face an aging workforce and decades‑old legacy systems, making it difficult to process critical applications quickly and cost‑effectively. Coordinating multiple synchronous and asynchronous systems with traditional RPA tools left gaps in speed, scalability, and business‑rule control. ​

After 24 years at IBM—and seven spent building AI‑automation platforms for U.S. federal clients—Naran began exploring multi‑agent frameworks in 2024. By 2025 his team had two CrewAI pilots running inside federal agencies, integrated with IBM’s WatsonX foundation‑model runtime. ​

CrewAI offered a hands‑on, open‑source framework that matured rapidly, let architects blend AI agents with custom code and business rules, and integrated cleanly with WatsonX. Strong support from the CrewAI team convinced IBM Consulting to pursue an enterprise license for multiple federal deals.

In implementation, Naran & team:​

• Built a hybrid “AI Flows + Crew Agents” architecture to control when rules‑based decisions versus autonomous agents run.
• Deployed multiple agents to extract applicant data from documents, summarize findings, and calculate program eligibility.
• Integrated CrewAI with WatsonX.AI to leverage IBM’s foundation‑model runtime while coordinating legacy and modern systems.

Early pilots show faster, more efficient eligibility determinations than legacy RPA—reducing manual coordination across disparate systems. IBM Consulting is finalizing enterprise‑license deals with federal agencies, aiming to move the pilots into full production and scale intelligent automation government‑wide.