For a while, ChatGPT was the default answer to any AI question. It was the thing everyone was using, the thing that made sense to try first, and for a lot of use cases, it was fine. We used it. We got value from it. But as AI became more embedded in how we work, fine stopped being enough. The switch to Claude wasn't impulsive. It came from using both seriously and noticing where the gap was. And switching is harder than it sounds.
“It was very difficult to move because ChatGPT has all your info, but I'd rather swap sooner than later and there are tools to export your history.”
Here's what actually drove the decision.
The difference in capability is real
This sounds like marketing language, but it isn't. When you're using AI to think, not just to generate text, but to work through a strategic problem, synthesise research, or reason about a decision, the quality of the output varies significantly between models. Claude handles complexity better. It follows longer chains of reasoning without losing the thread. It's more likely to catch its own errors and flag uncertainty rather than confidently fill gaps with plausible-sounding noise. For a consultancy, that matters. We're not using AI to write generic blog posts. We're using it to help clients think clearly about hard problems. The capability bar is higher, and Claude meets it more consistently.
It's what powers Artemis
We wrote recently about Artemis, our AI operator that handles research, content, and execution across our business. What we didn't spell out in that piece is that Artemis runs on Claude. That wasn't an arbitrary choice. The integrations, the reasoning quality, the ability to embed it into real workflows with real context, that's what made Claude the right foundation. Other tools are designed to keep you in their interface. Claude, through Anthropic's API and the ecosystem built around it, fits into the way we already work. That distinction is what separates AI as a novelty from AI as infrastructure.
It thinks, not just responds
There's a version of AI that's very good at producing output that looks right. Confident, fluent, structured. And there's a version that's actually reasoning. The distinction matters more than most people realise, because the first kind is dangerous at scale. You end up shipping things that sound good but don't hold up. Claude is more likely to push back, ask a clarifying question, or say it's not sure. That's not a weakness. That's what you want from something that's going to have real influence on the work you put in front of clients. We'd rather have a model that tells us when it's uncertain than one that papers over the gaps.
We believe in who we're building with
This one is harder to quantify but it's real. Anthropic's approach to AI development, the focus on safety, the published research, the way they talk about what they're building and why, aligns with how we think about technology. AI is going to be deeply embedded in how businesses operate. That means the values and incentives of the companies building it matter. We want to build with people who are thinking seriously about the long-term implications of what they're creating. That's not a universal standard in this industry. Anthropic clears it.
Switching isn't always the right move. ChatGPT is a capable product and there are contexts where it makes sense. But for the kind of work we do, embedded, integrated, high-stakes, Claude is the better foundation. It's not just that it performs better. It's that it fits the kind of AI operation we're trying to build.
