Craft Is Getting Drowned Out by Process

I began my design career focused more on process than on my skills. I built workflows for creating icons, including an entire language for how to construct them (our product had several hundred icons). When it came to actually designing, I often froze.

As I progressed in my career, I spent more time getting better at design and less time thinking about process. I used my anxiety as an advantage, staying up late learning about RIAs (rich internet apps) and RWD (responsive web design). I swung the other direction and kinda eschewed process altogether (down with wireframes!). I learned process doesn’t create great design, and it certainly doesn’t sell product.

AI has exposed this tension in a new way. As people learn how to use this new large-language-sea-change-inflection-point, craft is getting drowned out by process. People on LinkedIn share skill files and ways to connect systems. Others tout entire systems they build with AI (I built my own CRM while coaching my son’s T-ball team this afternoon!). Quality and craftsmanship are getting lost. On one side, the vibe-coded CRM you build will be rough: a UI that looks nice yet operates in bizarre ways. But even the usage of the product will produce lackluster results. As much as you’ll be wowed by AI’s first pass (dude, it just identified everybody I’ve emailed in the last fifty years with their telephone number, home address, and DNA sequence!), the results don’t stand up to scrutiny (oh wait, that’s a DNA sequence for a field mouse and the address has a ZIP code with an emoji in it).

This lack of quality isn’t a new take; we’ve been pointing out melting fingers in generative photos for a few years. My point is that while we all understand AI creates slop, we still oversell processes that generate these awful results. Or say things like “it got me 95% of the way there” without acknowledging the other 5% is like the wings on your airplane. And to get there, we spend so much time on process creation at the expense of quality results.

Building a business today has these issues times ten. Many businesses are being built around their “unique” processes. Marketing itself is full of agencies and products that promise what AI can do for content productivity or website performance. This era is supposed to be a boon for services companies, so it surprises me that more people aren’t focusing on results. We accept that AI creates lackluster results without people but then build an AI-centered business that ignores that reality. I’m building an AI-enabled business with my co-founder, Sara Croft, and I can say that very little of my team’s time is spent thinking about process and much more on quality.

My two decades of experience have taught me that the best processes are created retroactively rather than proactively. Do creative work, produce something valuable or compelling, then do it again. And again. And again. Somewhere along the way you start finding the parts you can build processes around.

This process of adopting AI makes me feel like I’m behind (Fear of Being Left Behind, or FOBLB). I spent all of 2026 building an AI-enabled process that generates Webflow sites from Figma, only to read weekly posts from people claiming one or both of those are dead and anyone who uses them should apply for unemployment soon.

Despite this FOBLB, I stick to my principles: I need to see my team produce strong work over and over again. I come in and fill in gaps with AI in ways that require almost no effort from my team. (That’s the benefit of AI; in the old days of 2018, to modify my team’s process required endless training and potentially adopting new products that many begrudgingly used.)

People keep paying our agency to build sites and design brands, and even more are starting to ask for deeper positioning and messaging work. How is that possible? I mean, these are things AI can do! The answer: quality and results. We do quality work. And we execute campaigns that create results. Sure, there is more AI involved in that today than there was in 2025, but AI is the supporting cast, not the main character. And we’re still moving way slower on AI adoption than outsiders would say we should be. But if we keep doing great work, people will keep paying us to do it. While people tear down their sites and rebuild in Next.js to save $35 a month with a site that looks identical to everyone else’s, we’re building fully custom sites at roughly the same speed.

The costs for us to do this were unsustainable last year. We were a new agency, so our prices needed to be lower to attract buyers. Some process improvement would create better margins, but mostly we planned for our profits to come from raising prices. Then AI blasted us all into orbit, and now these prices might have to stay the same to compete with the siren’s call of AI slop. So while the rest of the team has focused on creating effective marketing work and creative brands, I’ve kinda followed behind building AI tools in the gaps of the work. I made tools to generate OG images for blogs so my brand designer can build a richer brand. I made .md files to translate our messaging frameworks written in Google Docs into beautiful interactive pages so my marketers can focus on the copy. And lately I’ve been building our own client portals and dashboards to help our customers get more value without spending more money.

At some point, when I wasn’t lying awake with night sweats thinking about a LinkedIn post that claimed Webflow killed their dog, I worried that clients would question our use of AI and balk at paying anything when Claude Design could already create them a deck that uses italics for emphasizing phrases. But it turns out nobody brings it up. At all. Nobody cares. People liked our work and heard that it got other people results, so they were happy to pay.

Process doesn’t sell product. Process enables great product to come into existence, faster and cheaper. Many bytes have been spilled about how AI can generate content, but what it really helps with is process. And like overusing Gemini to create an oddly yellow “branded” graphic, overusing AI on process can create bland output and all the tools to measure performance without creating work worth measuring.