The AI-First Company: What Shopify's Viral Memo Means for Founders
"AI won't replace your job. But somebody using AI probably will."
That's the line Pat Walls opens with in a Starter Story episode I keep coming back to. It's about the memo Shopify's CEO sent to his entire company: the one that made the AI-first company a thing every founder now has to have an opinion on.
The short version of that memo: if you're not using AI at Shopify, you're falling behind on purpose. And before you ask for budget or headcount, prove AI can't do the job first.
Eighteen months later, that memo reads less like a hot take and more like a template. Duolingo went AI-first weeks after. Fiverr's CEO told his team "AI is coming for your jobs, mine included."
Here's the full episode, then my breakdown of what actually matters for you as a bootstrapped founder.
What Tobi Lutke Actually Wrote (Not the Headline Version)
In April 2025, Shopify CEO Tobi Lutke sent an internal memo to roughly 8,000 employees. It leaked, so he posted the whole thing on X himself.
Three lines did the damage:
- → "Reflexive AI usage is now a baseline expectation at Shopify."
- → AI has to be part of the GSD prototype phase (Shopify slang for "get stuff done"). You prototype with AI before you build anything.
- → Before asking for more headcount or resources, teams must demonstrate why they cannot get what they want done using AI.
He also added AI usage to performance reviews and peer reviews. Not a suggestion. A scored expectation.
Pat and Gus point out why this was such a shift. Two years earlier, the corporate default was the opposite: people got a slap on the wrist for logging into ChatGPT at work. Gus saw it at his own big tech job. Using AI meant you were "faking your work."
Now the CEO of a $100B+ public company was saying the reverse: not using it means you're not doing good work.
When someone on X asked Tobi what he actually uses AI for, his answer was "mostly internal stuff." Remember that. It matters more than the scary headline.
"Use AI or You're Fired" Is the Wrong Takeaway
The memo went viral a second time through spin.
Gus pulls up a LinkedIn post in the episode that claimed: "Shopify CEO to 8,100 employees: show me why AI can't replace you. 34% couldn't. They're gone."
That stat is nowhere in the memo. Yes, Shopify has had layoffs. But the memo never said "prove you're irreplaceable or you're fired." It said AI is now part of how work gets done here, and it gets measured.
The difference matters because the fear framing makes people freeze. The actual framing gives you a playbook.
Pat, who hires and fires at Starter Story, describes the two kinds of AI use he sees as a manager:
- → The lazy shortcut. Copy that reads like AI slop. He can tell instantly, and it's worse than not using AI at all.
- → The multiplier. Someone uses AI inside their workflow and the output comes back better and faster.
One gets you managed out. The other gets you promoted. Same tool.
And Gus is honest about where most people stand: 8 out of 10 people in his everyday life still treat AI as a silly toy. Inside his big tech employer, AI "still feels like a bad word." The majority is sitting this out.
That's not a threat. If you're reading a founder blog in your spare time, that's your edge.
The Pivot Table Moment (Why Normal People Get Left Behind)
My favorite analogy from the episode is boring on purpose.
Gus remembers showing coworkers a pivot table in Excel and watching them react like he'd done magic. He wasn't smarter than them. He just knew one feature they didn't, and it made him "the tools guy" at the company.
AI is the pivot table times 1,000.
Except this time the pivot table doesn't summarize the data. It does the work of the whole team that used to argue about the data in meetings.
Pat's advice for anyone who feels late: go look at how broken everything still is.
When they recorded the episode, ChatGPT's Operator agent couldn't get through a captcha. AI agents were mostly a promise, not a product. Pat's reaction wasn't disappointment. It was relief: "Oh, I'm so early to this right now. All I need to do is show up and I'm ahead of 99% of people."
Gus, who uses this stuff daily, says he's still using maybe 1% of what the tools can do.
The YouTube comparison lands hard here. Ten years ago people were asking "is it too late to start on YouTube?" Imagine saying yes to that. Same question, same wrong answer, new wave.
If you want to know which of your skills still matter once everyone has AI, I wrote about that in AI-proof skills. Spoiler: taste and distribution age well.
The GPT Wrapper Guy Who Got Promoted Instead of Fired
One story in the episode is the whole thesis in miniature.
Pat talked to a founder who built a GPT wrapper as a side business while holding a full-time job. His boss pulled him into a room. The guy was terrified. He was sure this was the "you're fired" conversation. (Pat went through the same moment when he started Starter Story while employed. Same fear, same room.)
Instead, the boss said: I saw what you built. It's awesome. Can you lead AI projects here? We need the help.
Think about what happened there. The employee taught himself to ship with AI on nights and weekends. The company had a CEO-level mandate to find AI wins and nobody internal who could execute. Supply met demand in a conference room.
That's the real meaning of the Shopify memo for individuals. Every company now has a standing bounty on people who can actually build with this stuff. You collect it either as a promotion inside the building or as a founder outside it.
The skills are identical. Automate a process, cut a cost, 10x a workflow. Do it for an employer and you get a raise. Do it for a niche market and you get MRR.
How to Run an AI-First Company When You're a Team of One
Most people reading the memo asked "will I get fired?" Founders should ask a better question: what does the headcount rule mean when the headcount is me?
Tobi's rule was: before asking for more people, prove AI can't do it. As a bootstrapper, steal that rule wholesale. Before your next hire, before your next contractor, before your next $500/month tool: can AI plus two hours of setup do this?
I watched Greg Isenberg make the adjacent point about product: the 20-engineer SaaS team is becoming 1 to 2 people plus AI. I broke that down in Is SaaS dead?. The same compression applies to your ops.
Here's what the memo looks like translated for a solo founder or a tiny team:
- → Make AI your prototype phase. First draft of every feature, landing page, email sequence: AI, then your taste on top. Never the other way around.
- → Apply the headcount test. Write down the next 3 roles you think you need. For each one, spend one honest afternoon trying to get 70% of it done with AI before you post the job.
- → Start with internal stuff. Tobi runs Shopify and uses AI "mostly internal" (support macros, research, reporting, meeting notes). Low risk, instant payback. You don't need an AI product. You need AI plumbing.
- → Timebox the tinkering. Pat's whole pitch is project-based: pick one specific thing and ship it in 12 days. A support bot, an onboarding email generator, a churn report that writes itself. One project beats 40 ChatGPT tabs.
The founders I interview on the podcast keep proving the point with real numbers. Solo devs at $35K MRR. Two-person teams doing what 15-person teams did in 2021. None of them are AI researchers. They just stopped treating AI like a toy about 18 months before their competitors did.
FAQ
What is an AI-first company?
An AI-first company treats AI usage as a baseline expectation, not an optional tool. Employees prototype with AI before building manually, AI proficiency shows up in performance reviews, and teams must prove AI can't do a job before hiring for it. Shopify's April 2025 memo is the reference example, followed by Duolingo and Fiverr.
Did Shopify actually fire employees who refused to use AI?
No. The viral "34% couldn't prove AI can't replace them, they're gone" post was spin, and that stat appears nowhere in the memo. What Tobi Lutke did do: make reflexive AI usage a baseline expectation, add it to performance and peer reviews, and require teams to justify headcount against AI first. Shopify's layoffs were separate.
How do I make my startup AI-first as a solo founder?
Adopt the headcount rule for yourself: before hiring anyone or buying a tool, spend an afternoon testing whether AI can do 70% of the job. Use AI for every first draft, then edit with your own taste. Start with internal work (support, research, reporting) where mistakes are cheap and the time savings are immediate.
Is it too late to learn AI tools in 2026?
No, and the broken parts are the proof. Agents still fail on real workflows, most employees at big companies barely touch the tools, and daily users admit they use a fraction of what's possible. Late is when everything works and everyone's fluent. We're not close. It's the "too late to start YouTube" question all over again.
Steal the Playbooks, Not Just the Memos
The memo tells you where work is going. It doesn't show you how founders are actually cashing in on it.
That's what the podcast is for. Every week I sit down with bootstrapped SaaS founders doing $100K to $10M a year and get the real numbers: what they automated, what they hired for anyway, what they'd never let AI touch.
