If you've been sitting through another AI webinar wondering why the demos look so smooth, and your own prompts keep returning lukewarm results, this episode's probably hitting a nerve. Sarah Shepard, our COO, and I have been using AI at StringCan for three and a half years now, and the people who sound comfortable with it didn't find a better prompt or a smarter course; they found a reason to use it that had nothing to do with work.
You'll come away with a concrete reason to stop starting at the office and start at home, a clearer picture of why over-committing to one LLM is setting you up for a rough quarter, and a way to stop agonizing about privacy in a way that still keeps you safe.
This post is based on Episode 57 of Revenue Rewired | The Real Reason AI Isn't Clicking for You, and the Easy Fix.
If you'd rather listen than read, find the full episode on Apple Podcasts, YouTube, Spotify, or Amazon. It's worth your time.
Why Did Starting Personal Make AI Finally Click for Our Team?
Sarah kicked this episode off by asking me to reflect on how I first started using AI, and my answer caught me off guard. It wasn't a business use case. It was a compound interest app for my 19 and 21-year-old daughters, built in Claude at the dinner table in about four or five minutes. My wife Rachel watched the whole thing unfold, asked a few questions, and then told me later that night she wanted to come home and use Claude on our family's actual expenses before she tried it on a single client in her bookkeeping business.
Anyone who's worked with me the last three-plus years knows I have a custom GPT named Zeke. He's my oldest AI companion, built right after ChatGPT went public, and at this point, he's embarrassingly well-trained. His sister Zoe handles my travel planning. Sarah gives me a hard time about it, and she's not wrong. The point isn't that Zeke is special. The point is that I built him to help me, not to help StringCan, which is why I still use him every day, three and a half years later.
When you start with something you actually care about, whether that's a trip to Sicily or a retirement calculator for your kids, the reps stack up without feeling like reps. By the time the business application shows up, the muscle's already there. Rachel's instinct to test Claude on our household before taking it into her client work is the pattern I see working every time, and the one I keep recommending to CEOs who tell me their team is stuck in AI-curious purgatory.
What Does "Don't Build on Sand" Actually Mean for a Revenue Team?
Sarah brought up a caveat in this episode that any operator needs to sit with. Don't build your revenue workflows around one tool's feature set. Even during the week we recorded, Anthropic shut off third-party tools inside Claude, and the OpenClaw community erupted. People had built whole businesses off custom prompts and agents inside one platform, and some of them watched pieces of that disappear overnight.
My own blind spot was that ChatGPT was working too well for me. I wasn't pushing myself to try anything else because Zeke and the stack I'd built around him felt like enough. Sarah pulled me out of that, and I've been getting into Claude seriously the last few weeks. The tradeoff isn't about favorites. It's about resilience. If ChatGPT disappeared tomorrow, my job would be harder, but I'd still have a working brain, a working process, and an alternative to lean on.
The mindset to aim for is AI fluency, not AI dependency. Your team's ability to move between tools is worth more than the quality of any single prompt library. If a CEO asked me what to do about their team's AI strategy right now, I'd say make sure at least two operators are fluent in at least two LLMs, and that nobody on your revenue team is single-threaded to one platform.
How Do You Pick the LLM That Actually Fits How You Think?
Sarah's framing on this was the line of the episode for me. She said ChatGPT, Claude, and Perplexity are not interchangeable, and your cognitive style probably matters more than the feature list. She's ChatGPT's least-ideal user, it turns out. She doesn't want a back-and-forth conversation; she wants a specific output, produced quickly, with a visual prototype if possible. Claude started giving her that, and the switch was immediate.
My own take is slightly different. I think about it tactically, which tool will get the thing done fastest? I spent a whole Sunday running prompts through ChatGPT, asking Claude to improve what ChatGPT produced, then pushing it back through ChatGPT again. It turned into something like a car negotiation, which I happen to enjoy. I've bought cars by pitting three dealers against each other and letting them undercut themselves. Running LLMs against each other has the same feel, and it's taught me more about each one than any comparison article ever did.
If you're choosing an LLM for yourself, try two or three before you commit to one. If you're choosing for a team, pick a primary and an alternate, and rotate people through both so nobody on the revenue side becomes a single-tool specialist.
Why Does Team Experimentation Move Faster Than Team Training?
One reason AI spread across StringCan this fast isn't that we ran a polished training program. It's because Sarah and I, and the rest of our leaders, shared what we were trying in real time. The stair-climber story where Sarah opened ChatGPT to help her come up with a fantasy football name. The dinner story where I pulled up Claude and built a compound interest calculator in front of my kids. The Sicily trip, where I used ChatGPT as a walking tour guide in front of churches we couldn't have named on our own. None of that was structured; all of it got reps.
The leaders of the companies we work with often ask us how to get their people using AI. My answer lately is to stop trying to make the first AI story a business story. Let it be a family budget, a health question, or a trip. Once the confidence's there, the business use cases surface on their own, and your team starts showing up to client meetings with ideas that didn't exist the week before.
FAQ
Q: What's a low-risk way to start using AI without worrying about privacy?
A: Sarah's filter is the one I'd borrow. If you wouldn't mind anyone seeing the information, you don't need to be nervous about putting it into an LLM. Start with the things that pass that test, like trip planning, movie picks, and a compound interest model for your kids. The confidence you build there transfers directly to the work decisions later.
Q: I've been using one LLM for a while. Is it a problem if I don't branch out?
A: It's not urgent yet, but it's worth addressing before a platform change forces you to. Anthropic removed third-party tools during the week we recorded this episode, and the OpenClaw community felt it immediately. The safer move is to have real familiarity with at least one backup tool before you need it.
Q: Which LLM is best for my team?
A: There isn't one answer, which sounds annoying, but it's the truth. Sarah needed Claude's prototyping. I'm still on ChatGPT for the bulk of my day. Perplexity is its own thing. Try each, pay attention to which one matches how you think, then pick a primary with a real alternate.
Q: How do I get my team to actually use AI instead of just talking about it?
A: Leaders share what they're trying, not the polished wins, the actual experiments. Rachel made the call at dinner to go home and build a family budget in Claude before she took it to her bookkeeping clients. That kind of story, told out loud, moves a team faster than any rollout deck.
Q: What's the single biggest mindset shift you'd recommend right now?
A: Stop trying to think of a business use case first. Find something you care about outside of work, use AI for that, then let the business applications show up on their own. It's the pattern I've watched work for the last three and a half years, on myself, on our leaders, and on the clients who've actually built AI into their revenue engine.
Ready to Build an AI-Fluent Revenue Team?
At StringCan, we spend a lot of time with marketing, sales, and revenue leaders who know AI should be part of their engine and can't quite get traction. The work we do is less about rolling out tools and more about helping teams build the kind of fluency where experimenting stops feeling risky. That's how you move from interesting demos to real pipeline impact.
If any of this sounds like your team, we'd love to talk. You can also listen to the full episode to hear the unedited version of this conversation, including the Sicily tour-guide story and the wine certification curriculum I built with Claude, both of which I think you'll enjoy. Subscribe to Revenue Rewired so the next one shows up without you having to think about it.
