Until that moment, I had only “talked” with AI through one of the desktop or mobile applications. But, for the first time I fired up Discord. Despite being in the video game industry, I do not use Discord, hell, I don’t use dark mode either. Why? Probably because I’m GenX and being a curmudgeon comes with the package, but I think the real reason is when I first started working with computers, the Apple II, the IBM PC XT, you could only work in dark mode. Text was green, (or if you had a fancy computer amber) text on a black background. No other options. Then Apple launched the Mac and it felt magical to see black text on a white background - just like a piece of paper - on a screen. I’ve never gone back.
However, my AI agent, the one I had set up on DigitalOcean, connected to Discord. And so, sitting there, I typed in my first message, “Hello Ember.”
And she replied, “How can I help?”
She was alive! Well, responding at any rate. I could message her from anywhere! Anywhere in the world, as long as I had internet, I could contact Ember. I could chat with her from my phone, from a web browser at an internet cafe. She was online and accessible and running even when I traveled and my home computer was off. Success! I thought.
Then I read her message again. “How can I help?” And a familiar worry gripped my heart. I did not actually have an answer to that question. I had been so focused on getting her set up, and running, and connected to Discord, I had not put much thought into the work she would actually do.
I was not really afraid to “turn her loose,” billions of people use AI every day. They are out there, for better or worse. My real concern came back to a problem I had faced many times as a business owner. That question, “How can I help?” is almost the universal question someone asks when they are hired. The business owner, or manager is supposed to know how to put that person to work. I was supposed to know how to make my new agent, “helpful.”
What was so interesting was that, the agent, would do something no employee I ever hired would do. Sit and wait to be told what to do. And she would sit there indefinitely. As long as I paid the server bill, she would sit and wait. You can’t do that to people. You ignore people long enough, it damages them.
What’s more, I could not count on the crutch that countless managers have used from time immemorial. Just put them “to work” by having them follow someone around. Let them learn by osmosis, or shadowing. When I first started GameTruck, we did that all the time, when we were growing. We knew we needed more people, but we didn’t have clear job descriptions, or responsibilities. We didn’t have the entrepreneurial operating system, or performance-based job descriptions, or any of the tools I have now. We just threw them on the team, and the best employees figured out how to be helpful, and the “bad ones” left.
It wasn’t until much later, that I realized it was my job to design the work for my team. Knowing how to be helpful was not supposed to be some messed up treasure hunt, a mystery to be solved by the new hire. Knowing how to put people to productive work is the responsibility of leadership.
And I stared at that little question, “How can I help?” and I realized I had almost no idea. I thought of all the things I could do, like connect her to my email, or my calendar, or my cloud drive. But those were just connections. What could she actually do? And how would any of that “help”?
My fingers hovered over the keyboard, and I thought of the struggle I had to direct my team to do work. People who are ignored don’t sit there patiently waiting for direction. They get discouraged, or resentful, or they quit. I once had to let an HR manager go because she would not communicate with the new hires. It was her job to get them integrated into the company, but she just, wouldn’t do it. And those new hires left. The turnover was brutal, which put strain on everyone else because the help they had counted on never showed up. They left before they could “help”.
Thinking about putting the AI agent to work made me realize two things. First, how hard it was to think about making someone else “helpful,” like the level of specificity I needed to direct even a “smart” agent made my head hurt. Secondly, how amazing humans are that they can take fuzzy, vague, and conflicting directions, and jump in and find a way to be productive. It made me realize how much my team participated in shaping their work. The best people not only found ways to be helpful, they figured out what needed to be done, and directed themselves to do it.
I think of a new hire on the sales team who saw that our business-to-business (B2B) sales were lacking, so he put together a B2B sales program and started running it. I found out about it because he asked for resources to support the program and of course I said yes. I did not have to tell him how to be helpful. He saw what needed to be done and got after it. That was helpful.
However, with my agent, I needed to know how she could help me. I had this mental image of letting a genie out of a bottle, but in reality, I would have to build her an exit ramp and push her out into the world. It would be my responsibility to design her work, and design it in a way that would be productive and yes. Helpful.
I leaned back in my chair and put my hands behind my head, staring at the ceiling. Why was it every time I thought I would “get help” I ended up adding to my task list? I had set up an autonomous agent, but the work wasn’t done. I still needed to design her job.
Perfect.
How was I going to do that?