Leadership can be lonely, even isolating. Whether you’re running a department or an entire company, you don’t always have someone to bounce ideas off, commiserate with or hold you accountable.
I’m fortunate to have both a business coach and a standing brainstorming session with other agency owners, but those are biweekly events and issues always come up at inopportune times.
That’s why I built a custom GPT named “Orby” to be my AI business coach. He — yes, I’ve anthropomorphized my AI — has helped me work through challenges in the office; identify patterns in our analytics and profit and loss statements; and develop new offerings.
Building your AI business coach requires a paid version no matter what Large Language Model you use. I built the original Orby on ChatGPT but recently rebuilt him on Claude. On ChatGPT, custom GPTs are called (ahem) CustomGPTs, on Claude they go by Projects, and on Gemini they’re Gems.
First steps
Custom GPTs are built with a set of instructions (a recipe, if you will) and knowledge docs (your ingredients.) Your instructions assign the GPT a role and purpose, specify tone and voice, provide context and basically tell the GPT how it should behave.
Writing out the instructions can be time-consuming and intimidating, especially if you’ve never done this before. ChatGPT offers a simpler approach. Once you’ve clicked on the “Create” button on the Explore GPTs page you can just describe the AI Assistant you want to build, and ChatGPT will write the instructions for you.
If you’re using Claude or Gemini, just describe the custom GPT you want to create and the AI will write the instructions. Then you can copy and paste those instructions into your new Project or Gem.
Your knowledge docs are the supporting information. My knowledge docs originally included information about flyte new media, my brand voice document, examples of my writing style and my vision of where I wanted to take my company. Over time I’ve added additional documents including transcripts of sales calls, our Brand Guide and summaries of my bi-weekly coaching sessions.
Once you’ve assembled your CustomGPT, Claude Project or Gemini Gem, you’ll want to take it out for a test drive or 10. You chat with it the same way you do with the regular LLM platform: by prompting it. Here are some sample prompts:
- Review this sales call transcript and provide feedback on what worked and what could be improved.
- Help me respond to a client who has been unresponsive to my project manager. Here are the details …
- Draft a thought-leadership post for LinkedIn in my voice about the recent changes in digital privacy laws.
Chances that what you build will be far from perfect. Building an effective AI business coach requires multiple rounds of iteration.
AI is very good at pattern matching, so providing examples of “good” and “bad” outputs — either in the instructions or in your knowledge docs — is critical to getting better outputs and making your GPT a more effective coach.
If you get responses that feel off, you’ll need to change either your instructions or update your knowledge docs. Keep iterating until you’re consistently satisfied with the outputs.
Magic mirror
Obviously, having an AI business coach is not the same as working a human business coach or having peers in a mastermind, and that human connection offers many benefits AI can never replicate.
Rather, talking to your GPT is more like talking to yourself in a mirror … except this mirror can respond, perform research on your behalf, provide fresh perspective and draft copy in your voice and on your behalf – and it’s available whenever you need it. Or him. Or her.
In fact, “talking” to your GPT isn’t metaphorical. The mobile versions of ChatGPT, Claude and Gemini allow you to hold verbal conversations with your AI business coach. This is like having a version of yourself available to chat 24/7 who can push back and push you to new heights. Your morning commute may never be the same.
