Every Founder Has Tried AI. Almost None of Them Operate With It.

I co-founded Omnidesk, a standing desk company in Singapore. We’ve done 8 figures in revenue. The business was running, but I was running it the hard way: buried in dashboards, firefighting marketing decisions, making calls that should have been delegated years ago.

I’ve been using AI since ChatGPT launched. Custom GPTs, prompt chains, the works. I could draft faster, brainstorm better, summarize anything. But I couldn’t make it run my business. After 18 months, I’d hit a ceiling.

When I switched to Claude last October, and then to Claude Code in January, the ceiling disappeared. Not gradually. Within weeks I was building systems that changed how the entire company operates.

The difference wasn’t intelligence; ChatGPT is smart. Claude Code could actually do things: read files, write scripts, connect to APIs, build tools that run on their own. It went from conversation partner to builder.

My daily CEO reporting went from 2 hours to 5 minutes. That’s not a rounded number for effect. I built a Claude-powered system that pulls data from Shopify, Klaviyo, and our internal tools, then delivers a morning briefing I can read in the time it takes to finish my coffee.

Our email marketing flows now convert at 34x the rate of bulk campaigns. Not because we hired an agency. We rebuilt our entire Klaviyo audience infrastructure (10 new segments, suppression lists, exclusion templates) in a single sprint, with Claude handling the segmentation logic.

We cut team reporting time by 80% across 7 direct reports. People who were spending half their Monday compiling dashboards now spend that time actually doing their jobs.

And the one that still makes me shake my head: our marketing attribution system exposed a 1,097x discrepancy between what Tiktok reported as ROAS and what actually showed up in our bank account. One thousand and ninety-seven times. That’s not a rounding error. That’s a fiction.

None of this required a technical co-founder, a dev team, or a six-figure software budget. It required a founder willing to sit down, learn the tools, and build.

I was a developer ages ago. I can read code and understand APIs, but I haven’t shipped a software product for a very long time. Everything I built was done through prompting, iteration, and breaking things until they worked.

I’ve been having the same conversation with other founder-CEOs. It usually starts with them asking what tools I use, and ends with them realizing the tools aren’t the hard part. The hard part is knowing what to automate, what to keep, and how to think about AI as an operator, not as a tech enthusiast.

After the fourth or fifth time walking someone through the same explanation over coffee, I realized I should just write it down. Not a polished case study. Not a sales pitch. The actual messy, specific, sometimes embarrassing details of what it looks like when a non-technical CEO decides to rebuild how his company runs.

Most of what gets published about AI for business falls into two camps.

This newsletter is neither.

Signal Before Noise is for founders who run real businesses and want to use AI to run them better. Not theoretically. With specific tools, real numbers, and honest assessments of what worked and what didn’t.

I’ll be writing about what I know: the operational layer. How decisions get made, how information flows, how a CEO stops being the ceiling of their own company.

The first series is the Omnidesk story: how an 8-figure standing desk company became an AI-augmented operation. What we rebuilt, the decisions behind it, and the specific results. No hand-waving. The actual systems, the logic, the prompts, and what happened after.

After that: playbooks, frameworks, and experiments. The playbooks are step-by-step builds for things like email segmentation and CEO daily briefings. Not theory. The actual setup, the prompts, the tools, and how long it took. The frameworks cover how I decide what to automate and what to keep manual, because that decision is where most founders get stuck. And the experiments are new tools and automations I’m building in real time, including the ones that don’t work.

I’ll also share what failed. The automations I abandoned. The tools that looked promising in a demo but fell apart in production. The Saturday I spent building a customer service bot that turned out to be worse than just having a human respond. The workflow that saved 10 hours a week until an API change broke it overnight. That part matters as much as the wins, because nobody else shares it.

I’m not going to pretend every tool I try is amazing or every system I build works on the first attempt. If something costs more time than it saves, I’ll say so. If a tool has a dealbreaker limitation, I’ll tell you before you waste a weekend on it.

Every post will have something you can take back to your own business. A number to benchmark against. A system to try. A question to ask yourself.

You’re a founder or CEO running a company doing $1M to $50M in revenue. You’ve tried ChatGPT a few times, maybe built a few automations, but haven’t figured out how to make AI a real part of how you operate. You’re tired of reading about AI from people who’ve never had to make payroll.

That’s who I’m writing for.

I’m not selling a course. I’m a founder who figured out some things and wants to share them with other founders. Here’s what I would have wanted to know 18 months ago: which AI tools actually work for operations (not just content generation), how to evaluate whether an automation is worth building, what the real time investment looks like, and where the limits are.

If you run a company, you know the problem. Dashboards, reports, Slack messages, ad platform metrics that may or may not be real. The volume of data has never been higher, and the clarity has never been lower.

The work I’ve been doing with AI isn’t about adding more tools to the stack. It’s about cutting through to what actually matters. Finding the signal before it gets buried.

That’s what I’ll try to do here, too.

What’s the one question about AI in your business that you haven’t been able to get a straight answer on? Reply and tell me. It might become a future post.