Creating Agentic AI Applications with Moveworks and lakeFS

Agentic AI is shifting enterprise AI from insight to action, enabling systems that can make decisions and execute across workflows and systems to drive real business outcomes.

Yet while 62% of organizations are experimenting with AI agents, only a small percentage have scaled them into production, revealing a significant gap between interest and execution.

Closing that gap requires a more disciplined approach to building and operationalizing agentic AI—it starts with the right data, governance, and architecture to support reliable, production- ready systems that can operate across real-world enterprise workflows.

Enterprise AI World recently held a webinar, Building Effective Agentic AI Applications, with Kevin McAbee, senior advisory solution consultant at Moveworks and Joe Pringle, VP of customer success at lakeFS, who discussed how to build data foundations that enable reliable autonomous decisions, implement governance and evaluation to drive measurable business impact, and more.

According to McAbee, effective AI begins with business outcomes. If an initiative is not tied to the business roadmap, it remains a science project. He recommends working backward from the desired result to find the processes, people, daily friction, systems, and actions that produce it.

Companies need to identify where AI benefits their use cases, McAbee said. High-value use cases are rarely mysterious. They appear wherever people compensate for disconnected information and workflows. For every use case, identify the data: what the agent must know; what it may access; and what it must be able to do. You need a unified command center to manage AI strategy, risk, performance, and value across the enterprise.

Pringle offered up the 4 pillars of a trusted foundation for agentic AI. This includes the following attributes:

• Isolated
• Reproducible
• Governed and Compliant
• Accessible

For the full webinar, featuring a more in-depth discussion, Q&A, and more, you can view an archived version of the webinar here.