The best implementations of AI I’ve seen to date value simplicity. There is one thing that people often forget when trying to get AI to do their work.
And that is valuing the ability to produce good results in ambiguity.
AI Agents, while they may appear to “think”, act in a world that is black and white. The prompts they would need to fully replace a human would be so astronomically large we may as well put the earth in a microwave and bake it to a crisp.
People however are inherent problem solvers and deal much better with exceptions and edge cases. When was the last day or week you had where everything ran 100% smoothly? Exactly.
So to cut to the chase, if you’re looking to get the best results out of an AI agent to free up your time, here’s what needs to be done.
Do it manually first.
Do it manually again, and again.
As you do it, document it, what goes right, what goes wrong, and what you did to fix the problem.
Then, put your documentation into your AI tool of choice (I use Claude).
Prompt it to analyse the workflow and produce for you a prompt for an AI project or agent for you to use to produce the desired outcome. (REMEMBER TO CLEARLY ARTICULATE THIS!)
Get it to analyse the prompt against the original input.
Refine.
Build and test.
Refine again.
When you’re happy, use it.
I’ve done this in as little as 30 minutes with people. It’s not a hard job, but requires discipline and not rushing.
Keep it simple! You could be saving 10+ hours a week in no time.
How did you like today's newsletter?
Enjoy this post, refer new subscribers and get the Time Flux Method - what I use to help people save 2+ hours a day.