stop asking AI to do that… try this instead
Ever since AI made it's appearance, I've been quietly side-eyeing every dramatic AI promise.
Not because I didn't see the potential. I did.
But as a fractional COO who primarily works with small businesses, my whole job is identifying consistent triggers, setting clearly defined next steps, and making sure it's documented. When my team and I are implementing automations in ClickUp or Zapier, we're very much in control of the flow. So when something new and shiny like AI Agents hit the industry offering to do things FOR me... my instinct was to test it quietly in the background and definitely NOT go running toward it.
Which is what I've been doing. Keeping AI firmly in testing mode, leveraged it in small ways for my own business. Let other people be the early adopters.
the tides have clearly shifted…
I feel like two things happened at once.
First, EVERY single tool in my arsenal now has a fully integrated AI component. And second, tech companies have really leaned into making itself more structured and accessible. Less "glorified search engine" and more "actual operational tool."
I started seeing places where it could genuinely reduce the mental load, without replacing the real skill - human thinking.
here's how i got started
After a deep dive comparison between ChatGPT and Claude, I made a full scale shift to Claude (feel free to comment if you need a tutorial on how to do this).
Then I opened Claude and asked it to help me figure out where it belonged in my workflow. Which is so obvious, but literally why not utilize Claude's built in knowledge base??
““Act as my AI expert and strategist and help me identify some ways I can integrate you in my daily workflow. Ask me clarifying questions about my tasks and tech stack so you can give the best suggestions.” (Add in a brief description of your deliverables and work before starting the chat.)”
That one prompt kicked off a whole system. Claude gave me custom prompt suggestions, project instructions to make it repeatable, AND an implementation plan. From there, it was just a matter of testing, flagging what wasn't working, and refining until the output was actually useful.
Make it repeatable
Tell it your goal, not just your task. The magic isn't "write this email." It's "here's what I'm trying to accomplish — ask me questions and help me think through this." That context has changed everything about the output I get.
Test it against your standards. If it's doing too much thinking for you, or creating MORE work to clean up — that's data. Bring the roadblocks back to Claude and have it refine the instructions until it hits right.
Build it into a repeatable system. Once you've got a prompt that works, save it. Turn it into a template. Make it something your team could eventually use without you in the middle.
With Claude, It's no longer just about the prompts. It's about clear use instructions that turn it from novelty to an essential tool in your business infrastructure.
AI is starting to fit HOW I operate my own business, and I'm learning fundamental and transformative ways that these systems can be setup for my clients as well.
your turn
Where are you in this? Still in testing mode, actively building, or completely overwhelmed by where to even start? Comment below - I'd love to know where my people are landing on this.
