Writing
The blog.
Essays on building, shipping, and selling AI systems — plus the occasional rant.
n8n vs Make vs Vapi: the real AI automation stack
A working comparison of n8n, Make, and Vapi for AI automation builders. What each tool actually does, where each one breaks in production, and which to pick per job.
Read essay →How to price AI automation services (why your first quote is always wrong)
How to price AI automation services without undercharging. The three mistakes every beginner makes, the sales-call script that fixed mine, and my actual 2026 rates.
Why most AI automation builders quit in 90 days
Most people learning AI automation quit inside ninety days. It is not the tools. It is a specific gap in weeks 8-12, and three moves that get you out of it.
The best niches for an AI automation agency, ranked by how fast clients pay
The best niches for an AI automation agency, ranked after two years of real client work. Which verticals close in under two weeks, which ones drag, and why.
How to learn AI automation without Python (and what to learn instead)
You do not need Python to build AI automations that ship. The skill that actually matters, the five things to learn first, and why Python is where beginners hide.
How I actually use Claude Code
Not the demo version. The weird, personal, file-first workflow I have settled into after a year of living inside the thing.
The demo-to-deploy gap
A working demo and a working system are not the same species. Most AI builders are optimizing for the wrong one.
Build the dashboard before the agent
You cannot fix what you cannot see. Observability is not the last thing you bolt on. It is the first thing you build.
The single-agent trap
"One agent that does everything" is not the goal. It is the bug. Here's the math on why, and what to do instead.
The LLM is not the bug
The scariest category of AI bug is the one where nothing throws. The output just quietly drifts. Here is how to catch it.
Most of your automation does not need an LLM
The hardest thing about being an AI consultant in 2026 is telling clients the AI is the wrong tool.
What I actually do on a consulting call
A transparent walkthrough of the first sixty minutes, what I ask, what I find, and why most calls end with the same three fixes.
Cold outbound is not dead. It's just been done by people who hate it.
The worst outbound you have ever received was automated by someone who did not care. The best outbound you will ever send is automated by someone who does. The difference is not the tools.
Voice agents are a product problem, not a model problem
The reason your voice agent feels bad on a call has almost nothing to do with the model and almost everything to do with four seconds you didn't design for.
How to price an AI project
If you are charging for hours, you are charging for the wrong thing. Here is what you are actually selling, and how to price it so both sides win.
Hiring your first AI engineer: don't hire a researcher
The person you need is the one who can ship a working system by Friday. It is not the person who can explain attention mechanisms at a whiteboard.
Why your n8n workflow breaks in production (and the three patterns that fix it)
Your n8n workflow ran fine in testing. In production it dies in mysterious ways on a Tuesday. Here are the three structural patterns behind almost every production failure, and the fix for each.
AI automation for med spas: the three workflows that book appointments without hiring staff
A med spa with two front-desk staff and a full schedule loses eight to twelve qualified leads per week to voicemail. Three specific workflows cover ninety percent of that leak. Here is what they are and how to build them.
How to audit an AI workflow you didn't build
Someone built a thing, left, and now it's your problem. No docs, no tests, a 400-node canvas. Here is the exact sequence I use to go from 'I have no idea what this does' to 'I know exactly what's wrong.'
The AI automation retainer: what it is, what it isn't, and how to price it
Most AI automation retainers are hourly billing with a nicer name. A real retainer is something different—and knowing the difference is why some builders charge $3k/month while others charge $75/hour for the same work.
When a voice agent is the wrong answer
Voice agents are the best demo in AI right now. They are also frequently the wrong tool. Here is the two-question test for knowing when to build one and when to build something faster and cheaper instead.
What separates a $500 AI project from a $5,000 one (it's not the model)
Model costs are rounding error. The real cost is in ten other places you are not pricing for. Here is where the 10x actually lives, and how to scope and charge accordingly.
AI automation for home service businesses: what works, what doesn't
Home services is one of the best niches in AI automation. Low tech bar, high call volume, genuine pain around missed leads. But three failure modes kill most deployments before they reach production. Here is the map.
The right way to learn AI automation (if you're starting from zero)
Most people start with Python tutorials and abandon them in two weeks. The right learning order is not what you think. Here is the 90-day path, the three tools you actually need, and the mistake that kills most beginners before they ever talk to a client.
Is this AI project worth building? A framework for founders
The question is not 'can AI do this?' It can do almost anything in a demo. The question is whether the ROI clears the real cost—integration time, maintenance, and the three months before it stops surprising you. Here is the two-by-two.
The pilot that never ends
Every enterprise AI pilot has the same failure mode, and it is not technical. Here is what actually kills them, and how to design a pilot that cannot die.
What is AI, actually
Everyone knows what AI is until you ask them to explain it. A practical answer for business owners, builders, and anyone who has been nodding along in meetings.
How AI changed in five years (and what it means for your business)
From GPT-3 to autonomous agents: a practitioner's account of what actually shifted, what's still hype, and why the window for early advantage is shorter than you think.
How AI is transforming real estate (and what's still hype)
What AI can actually do for a real estate agent or brokerage right now, what's marketing fluff, and the three workflows worth building first.
How AI is transforming law firms
Contract review, intake automation, and research summarization are real. AI writing legal arguments is not. A clear line between what's working and what will get you disbarred.
How AI is transforming e-commerce
Customer support automation, product description at scale, and personalization that actually works. Plus what most e-commerce AI pitches get wrong.
How AI is transforming restaurants
Phone calls, reservations, no-show follow-up, and review management: the restaurant workflows that AI handles well, and why most restaurant owners haven't touched it yet.
I built a fashion brand with $200 and a debit card
Before AI agents and automation consulting, I ran a streetwear label called Roses On Roses. What six press interviews, 805 orders, and one very real education in margins taught me about building anything from scratch.