Moving AI from Pilot to Production

Agencies have launched dozens of AI pilots – but many never make it into production. They stall for many reasons, including integration challenges, policy friction, unclear ROI, and the difficulty of embedding AI into real mission workflows.

It is relatively easy to set up a pilot program – target a particular process, make sure the datasets are clean and accessible to your AI tools, upgrade key hardware and software, and test AI-driven outcomes for quality results. But the same steps are arduous and time-consuming when scaling up; untrustworthy data, out-of-date or obsolete infrastructure, and poor identity control systems, to name a few factors, all must be addressed before an AI pilot can be scaled across the enterprise.

Learning Objectives:

• Outline the elements that were improved to start the pilot and assess how much upgrading and improvement is needed to expand the pilot across the agency
• Assess the talent available within the agency to implement an enterprise-wide AI program and the hiring needed to execute
• Delineate budget requirements and whether performance improvements can generate savings that can be applied to expand the program