For AI to become more useful over time, it has to learn from you by gradually understanding how you work. The more context it has about your preferences, priorities, and feedback, the more it can take on for you.
A lot of people get frustrated with having to correct the mistakes your AI agents or workflows make over and over again.
Before we start: if you are getting this email forwarded to you and are not yet part of the lab - you can join the tiny AI lab here.
The Challenge:
This week, pick one workflow that you already use with AI. It could be summarizing your email, writing content, planning meals, or something else you come back to regularly.
Run that workflow as you normally would.
When AI gets something wrong, don’t stop at fixing the output. Teach it what it should have learned from the mistake.
The prompt to get you started
Drop this into your AI of choice to get started:
You got part of this wrong [explain the mistake]. First, update the output based on my feedback. Then tell me what general rule you should learn from this so you perform better next time. Finally, tell me whether that rule belongs in this workflow, my project instructions, or your long-term memory, and explain why.
If the lesson should be permanent, save it.
For example, if you’re using AI to summarize your email and it keeps surfacing newsletters instead of messages that need a reply, don’t just correct today’s summary. Teach it what “important” means to you so it can apply that rule the next time it summarizes your inbox.
Too Easy:
Once you’ve refined a workflow and taught AI what a good result looks like, try splitting the work between two AIs.
The first AI completes the task. The second AI reviews it. Its job is to check that the instructions were followed, flag anything that doesn’t match your preferences, and identify improvements you should make to the workflow itself.
Over time, that reviewer becomes a way of continuously improving the system. Every mistake becomes a chance to update the workflow, refine the instructions, or add a new rule that makes the next run better.
This is the same idea behind many AI agent systems. Rather than asking one AI to do everything, different AIs take on different roles and improve the overall quality of the result.
Why this matters?
The best AI workflows improve through feedback.
Every correction is an opportunity to make the next interaction better. Over time, those small improvements compound. Instead of repeating the same instructions every time, you’re building an AI that understands how you work and can take on more of the job with out the feedback you end up fixing the same mistakes over and over or giving up on the workflow all together.
That’s it for this week.
Hit reply and tell me what you thought of the challenge this week.
I read every response, and your experience helps me shape these challenges. If you get stuck or have questions, reach me at [email protected]