My AI pill took some time to digest.
I didn't want AI writing for me, and I didn't want it doing my work.
But for as long as I can remember, when I found myself in the middle of monotonous, un-creative, "just gotta get this done" work, I'd feel this brain tickle — "there has to be a way to automate this".
A decade ago, I stumbled into low-code software. Tools like Airtable and Zapier let me automate away busywork and build systems around how I wanted to work.
Then AI hit. And I didn't get it — copy-pasting in and out of a chatbot?
Then coding models got good. Suddenly anybody could build software from scratch to do anything (including me).
I dove in. I spent months talking to Claude more than friends and family (still true, sadly).
I studied the online AI hype cycles. Turns out, lots of people were drinking the kool-aid. But with each week's fresh round of "This changes everything" hysteria, I had a realization — none of this was trickling down to real businesses.
Real businesses stand to gain the most from AI. They have lots of real people, doing lots of real, operationally-complex work, every day. They don't need (or want) to "change everything". Their software and processes work, but they see the writing on the wall — AI is coming for us all, whether we like it or not. It's "AI or die" for every business.
They also know that integrating AI into their business is going to be slow, painful, and expensive. Most of them never even got around to integrating automation tools like Zapier, Make, and n8n for the same reason.
Around the time I had my realization about this "real business" AI gap, the trend of "headless" software was emerging — the idea that people no longer need to click through an app, they just tell AI what they want, then the software and AI work together to make it happen behind the scenes.
This got me thinking — real businesses need AI, but they also want minimal change. They only want AI and automation in the right places, and their team doesn't want another app to learn. This headless concept actually makes a lot of sense here...
So I started sketching out a platform that could make this a reality:
- Connect any API, sync in data from all your disconnected systems, store it in a format AI can work with
- Build custom AI agents from scratch with skills, instructions, and access to any file or folder you upload
- Build AI automation workflows with real code so they can handle even the most complex business logic
- Set up a Slack app and lightweight "surface" interfaces for forms, portals, and dashboards - anywhere a human needs to click something
- And the cherry on top: design the whole thing to be driven by coding agents — tell Claude what you want, and it hooks it all up in minutes
I realized this platform actually serves 2 sides of a market:
- Companies with at least one somewhat-technical builder who can prompt a coding agent to build the AI automation the company needs
- Consultants who sell AI automation services to companies — the same "real businesses" who need AI, but don't want to integrate it themselves
Fast forward months of furious building, and Mug was born.
Now I'm on a mission to install AI automation into as many "real businesses" as I can.
If you've read this far into my missive, I sincerely appreciate it.
If you have any suggestions or feedback for Mug, please email me, message me on LinkedIn, or open an issue on GitHub. I read and respond to everything personally, and ship product updates daily.
—
