Ages ago I double-majored in geography and ended up learning about "sustainability" alongside my design education. A core part of that concept is environmental stewardship — we build more sustainably so we can be better stewards of our world decades from now.
Oddly, this is an apt framework for how I think about AI today.
The promise of AI is that you can spin up anything rapidly. Websites, logos, entire applications. We're in an experimentation phase and that's natural. But along the way, we need to be aware of the pollution it creates.
One-off tools are spinning out at an unprecedented pace. While diversity in SaaS is good, too much makes it hard on buyers. And with most AI apps making outlandish promises, it's harder to vet tools than ever. The landscape is polluted. Time is getting wasted everywhere, from people convinced their version of Salesforce is better, to people persuaded that commenting on LinkedIn posts will get them to $1M ARR.
In the 2010s, dev shops and marketing agencies built their own frameworks and CMS systems to speed up their work. The problem was it created lock-in. If the agency folded or a client outgrew the custom framework, they were stuck, and the money saved didn't last. I lived this firsthand when I had to get my last company's blog rebuilt after the custom CMS we'd bought became poorly maintained.
I think about this constantly when I use AI to supercharge our workflow, for our agency, and more importantly, for our clients.
If I were completely self-centered about margins, I'd build everything with AI: websites, interfaces, content. And I've watched the LinkedIn parade of sites made in hours and AI-native tools promising gorgeous results in a few prompts. But what's materialized are sites that remind me of Bootstrap in 2014, functional, fast, and completely devoid of soul. Soul and personality happen to be two things we get paid to protect.
We steward not just brand and GTM strategy for our clients. We have to do so in a way that's sustainable for them long after we're done. If we optimize for our own efficiency with AI tools, we put them in a bad position. So we stuck with Webflow, efficiency be damned.
Then the MCP revolution started.
MCPs have been released at a rapid pace this year, and our preferred AI provider has been adding integrations just as quickly. The result is the marriage of efficiency and quality we'd been waiting for. I can take Figma designs from our lead designer, move them through Figma's MCP, connect to Webflow's MCP, and export directly into production. It could be faster if it were straight to code, but this way, clients can hand it off to anyone who knows Webflow. If they re-platform, it's much easier from Webflow than something custom.
MCPs let us use tools the broader market already uses. They let products that excel at certain things keep building UIs that are easier than prompt-and-terminal management. They let us be good stewards.
The sustainability framework I learned in a geography class turned out to be the right mental model for AI, not because the technology demands it, but because our clients do.
