Agentic AI is quickly moving from buzzword to practice – but making it work in a large enterprise requires more than just new models or tools. It starts with a stable, scalable automation environment and a clear strategy for how AI and agents fit into it.
In our recent webinar, we explored this journey with Air France–KLM (AFKL): how their automation landscape evolved, why moving to the cloud became a key milestone, and what this change unlocks for Agentic AI.
What do we mean by Agentic AI?
Traditional automation behaves a lot like a fixed recipe: you define the inputs, the tools, and the exact steps – and the system reliably repeats them at scale.
During the webinar, Bálint Kulcsár (Senior Automation Consultant, BCA) used a short story from his grandmother’s kitchen to highlight the contrast. She doesn’t need a written recipe to cook a familiar dish – she knows the goal, understands her ingredients and tools, and adjusts on the fly when something changes.
Agentic AI aims for the same kind of flexibility. Instead of only following a predefined script, an agent can decide how to reach a goal: which tools to call, which workflows to trigger, and when to involve a human in the loop.
Why moving to the cloud was the turning point for AFKL
A key message from their story is that you can’t simply “bolt on” Agentic AI to any existing setup. To support agents at scale, AFKL first had to modernize its automation environment.
That meant moving from an on-premise setup – their familiar, but limited “home kitchen” – to a cloud-based UiPath SaaS environment that could support:
-
consistent, reliable operations across the group,
-
easier scaling of automations,
-
secure integration of AI capabilities,
-
experimentation with Agentic AI without compromising governance.
This cloud migration was not a side project. It was the central enabler that made their next steps toward Agentic AI possible.
Navigating a complex cloud journey
During the session, Bas Delvoix (Head of the AFKL Automation CoE) walked us through this multi-year transformation. It was far from a simp
le lift-and-shift.
AFKL had to manage:
-
legacy systems and historical constraints,
-
multilingual infrastructure and region-specific regulations,
-
strict security and compliance policies,
-
divergent architectural standards across different parts of the organization.
Each of these factors created its own “turbulence”. The migration succeeded because architects, security teams, developers, and RPA product teams aligned early and worked closely together throughout the process.
Today, AFKL operates a cloud-based UiPath SaaS environment that delivers higher reliability, scalability, and much greater flexibility for both automation and Agentic AI initiatives.
Getting ready for Agentic AI use cases
With the new environment in place, AFKL can now explore agentic use cases safely and incrementally.
Rather than trying to transform everything at once, they focus on high-value scenarios where:
-
there is ambiguity or variability in the work,
-
human judgment is important but repetitive,
-
better decision support can improve operations and customer experience.
In these areas, agents don’t replace people – they provide structure, context, and recommendations, giving teams more autonomy and more time to focus on coordination, safety, and value-added tasks.
Key takeaway – and what comes next
The main lesson from Part 1 of our webinar is clear: Agentic AI needs the right foundation, and for AFKL that foundation was the move to the cloud. A well-governed, cloud-based automation platform is not a “nice to have”, but a prerequisite for scaling AI and agents safely in a complex enterprise environment.
As Bas put it: “With Agentic AI onboard, we might reach destinations we never thought possible.”
Organizations that invest in this foundation now will be able to experiment faster, learn quicker, and capture the benefits of Agentic AI ahead of those who wait.