DESIGN LEADERSHIP

How can ChatGPT make you a better manager?

Rediscovering good old delegation skills

Recently I came across prompt engineering techniques for ChatGPT by Microsoft. And you know what? They suspiciously resembled something I was very familiar with. It’s so exciting and ironic that people had to invent artificial intelligence only to master one of the most human skills I can think of.

Microsoft’s prompt engineering techniques are exactly what one would do to delegate a task to a human.

Basically, delegation is the science and art of “human prompts” that help you get the job done with someone else’s hands. The better you describe the task, the faster and better result you’ll get. And vice versa — as a famous proverb perfectly puts it:

Garbage in, garbage out

I believe ChatGPT and other large language models (LLM), apart from their primary role of assistants, are also a good way to train delegation skills without the risk of upsetting actual team members. Although management has existed for decades, it doesn’t excite us like AI nowadays. Maybe, that’s why delegating work to other people feels like a burden, and not everyone has the patience to do it right.

Giving tasks to AI feels more exciting than working with other people.

So, let’s see how delegating techniques in management resemble the best practices of writing ChatGPT prompts.

1. Provide context

Giving enough context increases the relevance of a solution. Your team members can choose to do totally different things depending on what you share with them. The same with ChatGPT: you ask it to generate a ticket purchasing user flow for an app so that you don’t have to compose it from scratch, and you get a generic list of steps irrelevant to your case. But if you expand the request with the target market, a short description of a user persona, mobile platform, etc., the result will be of much higher quality. That’s why it’s essential to share more details than the task itself.

2. Set milestones

There are a thousand ways to accomplish any task. Senior team members and the not-yet-existing “human-level” AI probably won’t need much guidance. But modern LLMs’ capability is probably like the one of junior specialists who step-by-step guidance and regular checkpoints. That’s why instead of giving a week-long task, you’d probably split it into 5 parts and instruct a colleague every day in the morning. And what about ChatGPT? The same: first you ask it to find something, then to filter/sort information, then to compose something based on it.

3. Define the expected outcome

There is no ideal outcome; it’s always a trade-off between a set of often contradicting characteristics. Do you need it quick-and-dirty or polished? Should it be visionary or feasible? High-level or tangible? Or maybe a little bit of both? That’s why both in management and collaboration with ChatGPT, we need to frame the outcome clearly and define its quality attributes, i.e., how flexible, scalable, detailed, profound, structured, proven, refined, etc. it should be.

4. Give examples

An example is worth a thousand requirements. Not surprisingly, we use so many references in design work. When I wanted ChatGPT to generate some specific cases for me, it didn’t fit the requested structure, so in my following replies, I gave it feedback about the generated cases, pointing out which of them were relevant, and which weren’t and why. The same in people management: you want to check team members’ intermediate results to provide feedback early in the process and avoid people’s vain efforts. Or you provide those examples in advance so that your subordinate can “steal like a designer” and spend less time.

ChatGPT is a tool, not magic.

Summary

Delegation is an effort investment; that’s why many people prefer doing a task on their own because it will presumably be “faster” than explaining it to others. But this approach doesn’t scale well. Hopefully, on the wave of excitement about AI, we start paying more attention not only to giving tasks to machines but also to sharing work with other people.

👋 Hi there! I enjoy two things — sharing knowledge and drinking good coffee. If you like what you are reading now, feel free to support me via the “Buy Me a Coffee” platform ☕ Thank you! ❤️

• ✍️ Here are all my design articles on Medium.
• 💼 Let’s connect on Linkedin. Drop me a note if you want me to present the topic of this article at a conference or meetup.
• 🏰 Bored by design articles? Check my other blog about photography and overlooked architecture.