Next Steps

In this module, we covered:

Why context engineering matters and how generic AI assistants fail without it
Using Kestra's AI Copilot to generate and refine flows by describing inputs and goals
Implementing RAG to build AI responses using real data from your own sources
Building autonomous AI agents that use tools and make decisions dynamically
Designing multi-agent systems where specialised agents collaborate on complex tasks
Best practices for cost, security, observability, and production readiness

You now have the building blocks to use AI across the full workflow lifecycle — from generating flows faster, to answering questions from your data, to automating tasks that don't have a fixed sequence of steps.

Where to go from here

Experiment with different LLM providers. The flows in this module use Gemini, but Kestra's AI plugin supports other providers too. Try swapping the provider and comparing results.

Build custom agents for your data pipelines. Take a workflow you already run and ask: which parts involve decisions that depend on external data? Those are good candidates for an agent.

Explore Kestra Blueprints. The Blueprints library has pre-built workflow examples you can import and adapt, including AI and agent patterns.

Share your learnings. Join the Kestra Slack community to ask questions, share what you've built, and see what others are working on.

Kestra documentation

External resources

Homework

Complete the hands-on assignment to test your understanding of AI Copilot, RAG workflows, and agents in Kestra.

See: Module 3 Homework

← Best Practices | Back to module →