Picture this: I walk into a room filled with people from all walks of life, sitting in a circle and sharing their stories about AI addiction and how it's taken over their life; debilitated by the fear of AI taking over their job and rendering them useless. I take my seat and listen until it's my turn. I take a deep breath, pause, and then begin… "My name is AJ. I tell everyone I'm a Copilot Fan. I make my own Agents with Copilot Studio. I even have a blog about it. But what people don't know is that I actually use ChatGPT waaaay more than I use Copilot. If people find out, what will they think of me? Am I a bad person? Should I be banished from society? Will my friends still trust me? Will my family still love me? I can't live with the shame anymore!"
Ok, that was a tad dramatic. But I won't lie… I do find it amusing and occasionally bothersome that I prefer ChatGPT over Copilot for just about everything. I really only go to Copilot when I need to use the "Work" experience, or to use one of my Agents that I've plugged into Copilot.
Why, you ask? Because for researching information, for having a conversation, for pushing the limits, and more, I've found time and again that I get better outcomes with ChatGPT even though Copilot is powered by the same GPT-4o LLM. I think it has to do with the fine-tuning and guardrails Microsoft has put in place to make Copilot a safer experience with firmer, narrower boundaries. Far too often, I find that Copilot either doesn't want to engage or provide answers to some of my non-vanilla questions. And I'm not talking about anything illegal, inappropriate, or outside the Overton Window. It wouldn't be a stretch to say that 90%+ of my research and day-to-day conversational interaction happens with ChatGPT, whether it's via a web-browser or my mobile device. I even talk to it from time to time.

I spend a lot of time talking to people about the dangers of lurking around on consumer-grade conversational AI platforms and the security risks it raises for your organization. And how Copilot BizChat and Copilot for M365 are much safer because of their Enterprise Data Protection (EDP). So why do I preach what I don't practice?
The reality is that if you listen closely to what I say – or what most experts say – the trouble isn't with using a consumer grade platform like ChatGPT (or insert your favorite platform here). The trouble is when you go uploading, sharing or discussing your business data with a public LLM. That's when you run the risk of sensitive and proprietary data entering the public domain and either getting exposed or used to train an AI model or LLM.


As one expert who I was on an AI panel with recently pointed out, "it's okay to ASK ChatGPT whatever you want; the problem starts when you start TELLING ChatGPT things private things about you, your business or your clients." So, ASK but don't TELL. And that my friends is the cardinal rule I live by when it comes to public AI platforms.
The moral of the story is that there's nothing wrong with "living on the wild side." Just be careful that you know where the red lines are drawn and you steer clear of them. So go play with ChatGPT, Gemini, Claude, Perplexity or whichever platform / tool rocks your boat. And it's ok to love them more than you love Copilot. But when it's time to work with your business data, come back home to the safety of Copilot for Microsoft 365.
Regardless of which platform you choose to use, you'll want to sharpen your prompting skills if you want the best experience and outcomes! Check out this post on writing great prompts.





