Everyone’s automating with AI now. Most of them are doing it wrong.
They’re spending four hours a day “prompt engineering” to save thirty minutes of work. They’re paying $600/month for tools that duplicate what three free ones already do. They’re building Rube Goldberg machines when they need a fucking hammer.
I learned how to automate my workflow with AI the same way I learned to code — by breaking things until they worked. No bootcamp. No certification. Just a laptop, spite, and a burning refusal to spend another year watching credentialed idiots get promoted while I did their work.
Here’s what actually matters.
The Automation Hierarchy Nobody Teaches You
Before you touch a single AI tool, you need to understand this: not all automation is created equal.
There’s a hierarchy, and most people start at the wrong end.
Tier 1: Eliminate the work entirely. Stop doing the thing. Say no. Delete the meeting. Kill the report nobody reads. AI can’t help you here — this is about growing a spine.
Tier 2: Systemize without AI first. Templates, keyboard shortcuts, text expanders. If you can’t systematize it manually, AI won’t save you. It’ll just generate garbage faster.
Tier 3: Now bring in AI. Once you’ve eliminated what doesn’t matter and systematized what does, AI becomes a force multiplier instead of a expensive distraction.
Most people skip straight to Tier 3 and wonder why they’re still drowning.
The Stack That Actually Works (May 2026 Edition)
Forget the listicles. Here’s what I actually use, every single day, to automate work that used to eat my life.
The Foundation Layer
Anthropic’s Claude 3.7 Opus — $20/month. This is your workhorse. Best reasoning, best at following complex instructions, doesn’t hallucinate itself into uselessness like the others. I use it for code review, documentation, and anything where being wrong is expensive.
ChatGPT Pro — $200/month. Yes, it’s obscene. But GPT-5’s deep research mode has saved me 20+ hours of technical research in the past month alone. Cancel Netflix. Cancel your gym membership you don’t use. This pays for itself.
Perplexity Pro — $20/month. For when you need citations and Claude’s training cutoff is screwing you. The API access alone is worth it.
The Automation Layer
Make.com — Starts free, scales to ~$30/month. This is where the magic happens. Zapier is for people who enjoy paying extra to feel enterprise. Make gives you the same power for a fraction of the cost.
Cursor — $20/month. If you write any code at all, this is non-negotiable. It’s VS Code on steroids, with AI that actually understands your codebase. I’ve watched it refactor entire modules while I made coffee.
Notion AI — Already included if you use Notion. Underrated as hell. I use it to auto-generate meeting summaries, project documentation, and weekly reports. Three minutes of editing beats three hours of writing from scratch.
The Specialty Tools
ElevenLabs — $22/month for the Creator plan. Voice cloning that doesn’t sound like a robot having a stroke. I use it for video voiceovers and podcast editing. Controversial take: Yes, it’s weird. It’s also 10x faster than recording everything yourself.
Midjourney V7 — $30/month. For when you need visuals and can’t afford to wait for a designer or learn to design yourself. The people who say “AI art isn’t real art” aren’t on deadlines.
How to Actually Automate Your Workflow with AI (The Framework)
Here’s the process I use. No fluff. Just the steps.
Step 1: Audit Your Hell Week
Track one week of work. Every task. Every interruption. Every “this will just take five minutes” that took an hour.
Write it down. Spreadsheet, notebook, doesn’t matter. Just capture the reality of where your time actually goes versus where you think it goes.
You’ll be disgusted. Good. Use that.
Step 2: Identify the Repeating Nightmares
Look for patterns. What are you doing more than twice a week that makes you want to fake your own death?
For me it was:
– Reformatting client data from PDFs into structured formats
– Writing similar-but-different email responses
– Generating boilerplate code for API integrations
– Creating social content from longer articles
Your list will be different. That’s the point.
Step 3: Build One Automation at a Time
Don’t try to automate everything at once. You’ll build a house of cards and hate yourself.
Pick ONE task. The most painful one. The one that shows up three times a week and steals an hour each time.
Here’s my framework for how to automate your workflow with AI for any repeating task:
Map the inputs and outputs. What goes in? What needs to come out? Be specific. “Make it better” isn’t a spec.
Choose your tool. Claude for thinking, GPT for research, Make.com for connecting things, Cursor for code. Match the tool to the job.
Build the dumbest version first. No fancy logic. No edge cases. Just make it work once, manually, with AI assistance.
Test it five times. Does it work? Does it actually save time? Or did you just spend six hours automating a 20-minute task? (We’ve all done this. Learn to recognize it.)
Systematize it. Turn it into a template, a saved workflow, a keyboard shortcut. Make running it cost you zero mental effort.
Step 4: Stack Your Wins
Once one automation works, add another. Then another.
I’m currently running 23 different AI-powered automations. They save me approximately 15-20 hours per week. That’s not an exaggeration — I tracked it.
But I didn’t build 23 automations in a month. I built one per week for six months. Then I refined them. Then I stacked them.
The Controversial Truth About AI Automation
Here’s what’s going to piss people off: most knowledge work doesn’t need a human doing it anymore.
Not all of it. But most of it.
The emails. The reports. The summaries. The reformatting. The “quick call to sync.” The status updates nobody reads. The documentation everyone ignores.
AI can do this stuff now. Not perfectly. But at 80% quality in 5% of the time.
And here’s the thing nobody wants to say out loud: if your entire job is tasks that AI can automate, you don’t have a skills problem. You have a job that shouldn’t exist anymore.
That’s not a moral judgment. It’s market reality.
The move isn’t to fight automation. It’s to automate yourself out of the bullshit so you can focus on the work that actually requires a human brain.
Strategy. Creativity. Relationships. Judgment calls. The stuff that’s still hard in 2026.
Let AI handle the rest.
The Real Cost of Not Automating
Time is the only asset you can’t get back.
Every hour you spend on work that could be automated is an hour you’re not spending on:
– Building skills that actually compound
– Creating things that matter
– Getting paid what you’re worth instead of what your current workflow caps you at
I spent three years doing everything manually because I thought that was what “hard work” looked like.
It wasn’t hard work. It was stupid work. And it kept me broke and tired while people who knew how to leverage tools lapped me.
Don’t make my mistake.
Your Next Move
You don’t need permission to start automating. You don’t need a course. You don’t need to wait until you “understand AI better.”
You need to pick one soul-crushing task you do every week and figure out how to automate your workflow with AI for that single thing.
Start tonight. Not Monday. Tonight.
Open Claude. Describe the task. Ask it to help you break down the automation into steps. Then build the first version.
It’ll be messy. That’s fine. Messy and automated beats pristine and manual every single time.
The people who win in the next five years aren’t the ones with the best degrees or the most experience. They’re the ones who figured out how to multiply their output while everyone else was still doing things the “right way.”
Be one of them.