Events Calendar

Presenter: Matt Houseman and Michael Emerick

Topic: Managing the "Messy" Parts of AI Development

For the May session, we’re planning to do a deep dive into what it actually looks like to build a tool using AI as a teammate. Most people use AI to write snippets of code, but we’ve been working on a framework, using Agentic AI and Microsoft Copilot, to manage the entire lifecycle of a project while keeping things clean and controlled.

The Project: From Unstructured Data to Actual Insights

We’ll be using a tool we built as our case study: an application that takes a pile of unstructured invoices and turns them into organized data and spend observations.

Instead of just showing the finished product, we want to walk through how it actually came together in less than two hours:

• The Blueprint:How we used "Procedural Guidelines" to keep the AI from going off the rails during the roadmap and technical design phases.
• The Workflow:A look at how we move from a high-level problem into specific user stories and prompts.
• Being Honest About Bugs:We’ll show the points where the AI tripped up—like a side-nav bar overlapping the content or multi-file upload glitches—and how we used those moments to refine our prompts and fix the code live.

Let’s Build Something Live

Once we’ve shown the "how-to," we want to hand the keys to the audience. We’re planning to:

• Take a live request:We'll ask the group for a new feature or an "edge case"—maybe something like predicting future spend or flagging duplicate payments—and try to implement it on the spot using the framework.
• Brainstorming Other Workflows:This approach isn't just for developers. We want to hear what kind of "messy" manual processes the group is dealing with in their own roles so we can talk through how to apply this same structured AI approach to solve them.

Our goal is for everyone to walk away with a practical mental model of how to use AI as a reliable, predictable collaborator for any complex project.