The article argues that AI workflows lack true competitive moats, emphasizing that value lies in business insights, accumulated knowledge, and client relationships rather than easily replicable prompts or tools.
Someone pitched me on partnering last week. He had a "proprietary AI system" for a niche and wanted me to resell it. I watched the demo and told him I could rebuild it in an afternoon with Claude Code and that’s how how he had built it too. So I asked what stops a competitor doing the same thing and he said the prompts. Quick context on why I care… I run an AI agency and we find the one expensive, repetitive process inside a business and build a system around it so the company doesn't have to hire an AI team. Anyone with the same tools can copy every workflow I've ever shipped, including the ones I charge five figures for. I've made peace with that because the workflow was never really what they were paying for. What they're paying for is that ik where the breakage is. One client was losing somewhere around 20 hours a week matching invoices to purchase orders. The cause turned out to be a single supplier who puts the PO number in the email subject and nowhere else. I found it by sitting in their office for 3 weeks watching someone do the job. A competitor can copy the workflow but they can't copy those three weeks The other thing you can't copy is the year after. Version 1 works fine on demo data. In production it hit a couple 100 exceptions in the first month and 12 months on the system is maybe 20% AI and 80% accumulated notes on how this one business behaves. Supplier formats that change without warning, approvals that happen over WhatsApp instead of email, that kind of thing. None of it is in the prompt and most of it isn't written down anywhere except in my head and a very long Notion page. When the software breaks at 2am, someone gets called. Businesses pay for that person to exist and the software is almost incidental. Clients say they could get it cheaper but they still use it because last time it broke I'd fixed it before their team noticed. The one I didn't expect to matter this much is distribution inside a niche. Client one introduces you to client two. What isn't a moat: your prompts, your model choice, your tool stack, anything you'd put on a slide with a lock icon next to it. All of that is an afternoon away from anyone who wants it and the models change underneath it every quarter anyway. So if you're an agency, stop selling workflows. Sell the fact that you know where their money is leaking and that you'll pick up the phone when it breaks. Your moat has to be data you accumulate or integrations that hurt to rip out or distribution you own. Most of us don't have a moat the way investors mean the word. We have a head start and a relationship. That's fine, it's been the agency model since long before AI. The mistake is pretending the afternoon is the asset and pricing it like one. TLDR: anyone with a coding agent can copy your AI workflow in an afternoon, mine included. What holds up is knowing where a business is losing money, a year of its edge cases, being the one who gets called when it breaks, and referrals inside one niche. Prompts and tool stacks aren't a moat. Price the relationship and not the afternoon.
A reflection on whether the competitive advantage in AI is shifting from the models themselves to the workflows and user experience built around them, using examples from coding tools and note-taking apps.
The article argues that abundant AI intelligence will not eliminate business moats but will shift value to companies that translate AI models into real-world outcomes through coordination, workflow data, and structural necessity.
A founder shares his experience with AI tool adoption, noting that most people collect tools without achieving real results. He advocates focusing on one critical business problem and iterating until the workflow genuinely works, citing his own success reducing client reporting time from 4-5 hours to under 45 minutes.
The article argues that AI companies' competitive moat, built on expensive model training, is easily undermined by distillation—replicating models through repeated API queries—as demonstrated by industry practices like xAI training Grok on OpenAI models and Anthropic accusing Chinese labs of mining Claude.