You hired AI. But did you onboard it?

AI is so accessible that it’s easy to assume we can give an LLM a task and expect it to produce exactly what we have in mind.

But many of us use AI to replace our workflows without even first explaining them step by step, or what a good result should look like. What happens then? The result is just more output, not necessarily what you expected.

We then spend time tweaking prompts and telling ChatGPT what to change. Eventually, it comes up with something acceptable, and we use it... not because it’s exactly what we wanted, but because we aren’t sure how to guide it to make what we want.

Imagine a company hiring someone new and immediately expecting them to produce the same work as Jeremy, who has been there for six years. The new employee has no context, no guidelines, no standards, no business awareness...

AI is no different. Before expecting strong results, we should give it the relevant context, guidelines and examples.

How do you “onboard” AI? Do you write instructions and examples with every task, or use another approach? And has this improved your results?