Covered in This Course
- Design an AI agent for a specific finance task by identifying which workflow steps should be rule-based and which should rely on AI judgment.
- Develop reusable AI skill files that produce the same work the same way, every time.
Description
AI agents are opening new possibilities for how finance teams work, but building an effective agent requires more than a good prompt. In this hands-on course, you'll start with a prompting framework built for finance work, then learn how to design and build AI agents around real workflows.
You'll explore how to break a process into inputs, decisions and actions, determine where AI judgment adds value, and identify where traditional formulas, rules and controls are still the better choice. You'll put these concepts into practice by building your own finance agents from start to finish and learn how to recognize when a workflow has become complex enough to require multiple coordinated agents.Whether you're beginning to experiment with AI agents or ready to move from prompting to building, this course gives you a practical foundation for creating solutions that are useful, reliable and designed for finance.
This is the first course in a three-part Finance Agent series. It will prepare attendees to participate in the AFP Hackathon at the 2027 AFP FP&A Forum in New Orleans this March.
This educational badge series uses the following definition of an agent: an AI model equipped with tools and guided by instructions and guardrails that allow it to perform a defined job or workflow. This aligns with definitions from frontier models, although there is a bit of wiggle room for differences.
The labs start deliberately with human-guided agents rather than full autonomy. Keeping a person in the loop makes it easier to learn how an agent works, see where it succeeds or fails, and practice the judgment and governance needed to use it responsibly. Think of this as the foundation, not the finish line. At the FP&A Forum, we will build on that foundation and explore what it takes to let agents run more autonomously, and when finance should and shouldn't let them.