I Started a Podcast at Work

I haven't been blogging as much lately because my creative energy has been focused on creating a podcast that I share internally at work. In this post, I'll discuss how I came up with the idea, the format of the podcast, and some reflections on it (spoiler alert: it's fun!).

How It Started

AI hype took off this year with the release of Cursor, Windsurf, and the like. I worked with a few colleagues to develop our Engineering AI Strategy, and as I did that, I realized there was so much to learn about working with AI in software development. I learned the most in one-on-one chats with people about their work.

So I kicked off a "podcast." Scare quotes here because it is just a 15-minute recorded meeting. The only editing I do is sometimes add a tacky title and trim off the beginning and the end. The rest is a casual conversation about AI workflows.

The Format

The format of the show is straightforward. We have over 500 people in product, design, and engineering, and their day-to-day work varies widely. So we start with an intro that describes where in the organization the person works, talk about the AI tools available to them, and then talk through some examples.

What I've Learned

• Tools vary widely across the organization. Even in engineering, AI tools that are good at TypeScript are not good at Kotlin.
• There is a common evolution of individual workflows:
  • Start by asking questions about the specific code you are looking at.
    • Experiment with "agentic" workflows by asking the AI to "do the thing" for you.
    • Get frustrated that agentic workflows generate too much code.
    • Try again, but give the agent all the context you already have.
    • Have an "Oh, sh*t" moment as the agent does the task for you.
    • Give the agent a big task and have an "Oh, no" moment as it messes up.
    • Finally, build a workflow that uses the agent as a pair programmer. Break tasks down with the agent. Develop markdown to-do lists, start a new chat, and iterate through the to-dos one at a time.
• Not related to AI: People love talking about their work. And I love listening.

Why You Should Start a Podcast at Work

Aside from helping people share what they've learned about AI, this podcast has been an excellent opportunity to get to know my co-workers and help share and celebrate work across the organization. I've learned about business challenges in different teams, technical problems that people are solving, and new ways of approaching problems.

I've learned all that while helping people learn from their coworkers and share their experiences.