I see it all the time on tech forums and social media: smart, experienced people complaining that AI models and agents just don’t work for them.
This always surprises me, because my experience has been the exact opposite. I feel like I’m getting great value from these tools, especially in automating mundane tasks. A good chunk of the mechanical part of my job—typing words and symbols into a keyboard—could be delegated. And while the AI might be typing those symbols and words, the thoughts, the engineering, and the analysis of the trade-offs are still mine.
So, it did got me thinking, why the disconnect? Is my work more mundane than I think? Am I just getting lazy?
While we could debate those possibilities, I realized the answer is something else entirely. It clicked when I recognized a pattern: I’ve been communicating with AI in the same way I’ve mentored early-career professionals for years.
Before I delegate any task, I follow a simple playbook:
- Envision the End Result: I get crystal clear on what “done” looks like.
- Break It Down: I split the task into smaller, manageable steps or goals.
- Establish Rules & Guardrails: I provide clear instructions (“use this library,” “follow this format,” “use X as an example”) and set firm boundaries (“don’t use this outdated method,” “avoid discussing Y”).
In the past, the output of this process was detailed Jira tickets or GitHub issues with tasks and sub-tasks. Today, it’s a set of well-crafted prompts.
Of course, I’m not equating people with machines. But the approach of clear, structured delegation that works so well in mentorship has been incredibly effective for me in this new AI-driven world. It’s a mindset that transforms the tool from a frustrating, overly optimistic, black box into a result-driven assistant.
