While an AI pilot project can help demonstrate a use case’s feasibility, it can also make full deployment seem too easy. That’s true of many kinds of IT pilots, but even more so with AI, where resource demands and management challenges increase significantly as projects scale. The problem is that a project that proved feasible in the pilot phase might not be worth the additional investments needed to deploy at scale.
To change this equation, agencies need to create an underlying infrastructure that supports a wide range of enterprise AI initiatives. That includes the ability to manage both large language models and AI agents across their lifecycles, so the software is always up to date and complies with audit, zero-trust and other governance policies. That allows an agency to make the leap from a limited pilot (e.g., one developer fielding three or four agents) to enterprise-level deployment (AI-augmented development processes).
“There’s a big disconnect between a project running on a developer’s laptop and an actual production-ready implementation,” said Adam Clater, a Chief Architect in the Field CTO Organization at Red Hat. “And it’s in bridging that gap that I think a lot of agencies are starting to find some difficulties.”
In this video interview, Clater discusses how agencies can advance their AI maturity. Topics include:
- Factors to consider when scaling to operational deployment
- The benefits of an AI-assisted development workflow
- How LLMs might be able to leverage existing application programming interfaces