MI.NET - Dustin Kingen - Teaching AI to Use the Computer
Details
Agenda
- 5:30 - 6:30pm Networking
- 6:30 - 7:30pm Presentation
- 7:30 - 7:45pm Closing
Abstract
Developers are increasingly building agents into real software workflows. As those systems become more capable, the engineering challenge is no longer simply getting an agent to perform a task. It is designing the environment around it: deciding where the agent should exercise judgment, where the system should remain deterministic, and how those pieces work together reliably.
In this session, we’ll explore how to design systems where agents can productively use the computer as part of a larger workflow. We’ll focus on how to define the boundaries around agentic behavior, constrain the uncertainty it introduces, and build the surrounding software so the system remains understandable and dependable.
Through practical examples, we’ll look at how one-off AI-assisted tasks can evolve into repeatable engineered workflows. We’ll examine how to recognize what is worth automating, make use of capabilities that already exist, and identify where additional tools or infrastructure are actually needed.
This session is language-agnostic and focused on practical engineering concepts rather than a particular framework. Attendees should be comfortable using AI assistants or agent tools such as ChatGPT, Claude Code, or Copilot, and have some familiarity with automation through scripts, APIs, or similar tools. No prior experience building AI agents is required.
Bio
Dustin Kingen is a consultant and software engineer with 15 years of experience in enterprise systems. He organizes MI .NET and MITechCon, and speaks on modern architecture and authentication. Known for his first-principles approach, he is the person teams turn to when solutions aren’t on Stack Overflow.
