✏️ Note: Originally shared in draft. Took less time to write than to finally hit publish. Classic.
Shopify just dropped the AI mic two weeks ago with a letter to their entire company. Its lessons are relevant for all of us. Frankly speaking, if you’re in the workforce and not already into AI, this memo is your wake-up call. It’s an invitation start building with intent. The shift — from tinkering to transformation. Kids, playtime is over.
I read the letter once. Then again. Then I decided to upgrade my sprint playbook. For its full content, check out Tobi’s tweet here. (Yes, I still call them tweets. X can’t take that from me.)
In this post, I break down why this matters, and how to actually use it after you finish reading the memo. It’s split into three parts to ease you in:
Section-I: Warm-Up | AI Isn’t the New Electricity — It’s the Oxygen
Section-II: Evolve or Expire | Why most companies won’t survive AI if they don’t adapt
Section-III: Key Lessons | 6 shifts from Shopify’s playbook and how to use them today.
Tobi’s memo isn’t just smart — it’s a blueprint for modern organizations.
I’ve spent years in this space. And this memo nailed in a few pages what most panels struggle to explain in hours. I’ll be sharing this post whenever I hear someone says “AI is just hype!” again.
(Yes, I feel like Amy from Big Bang Theory — “Please pass the butter!”)
If it’s not embedded in your team’s daily habits, you’re behind. We’re no longer operating on fixed capabilities — every other week, machines surprise us. The bar keeps rising.
Smart companies treat AI like the first draft of every workflow, not a feature add-on. But here’s the catch: it’s a chicken-and-egg problem. Most people aren’t equipped to even ask that question. High-AIQ employees just get it. Everyone else depends on those who do.
The edge isn’t just access to models , it’s how fast your team makes AI second nature. Agentic AI is moving at breakneck speed, but most of the workforce hasn’t caught up. Inertia sets in. And when teams chase unrealistic goals, broken systems take over: misaligned KPIs, burnout, and shallow adoption.
Worse, in the rush to “do AI,” teams often overlook security risks and ethical blind spots — hallucinations, shadow deployments, or leaking sensitive data via prompts. Leveling up is necessary, but not at the cost of burning people out or compromising trust.
It’s the strangest, most powerful wave we’ve ever surfed as humans. We’d better learn how to ride it. Fast.
Just two weeks ago, Tobi Lütke — co-founder and CEO of Shopify — shared an internal memo with the entire company. On the surface, it’s about AI. But really?
It’s about how to redesign the way we work — from the ground up.
For some context: Tobi Lütke has made it clear that Shopify needs to lean into AI to keep up with its rapid growth which is 20–40% year over year. But here’s the kicker: they’re not planning to grow the team to match. In fact, while revenue has climbed, headcount has gone the other way — from 11,600 employees in 2022 down to just 8,100 by the end of 2024. It’s a company laser-focused on doing more with less.
It is an absolute goldmine for anyone building, leading, or just trying to stay relevant. Because here’s the truth: AI-native companies aren’t inherently a tech problem. They’re a design problem.
This shift isn’t just for engineers. It’s for ops. Marketing. Support. Everyone.
AI isn’t a tool to adopt — it’s a foundation to build around.
This isn’t a trend. It’s a blueprint.
And Shopify’s memo gives us a clear view of what an AI-first company really looks like:
Here’s the punchline for builders:
Your productivity and effectiveness will hinge on how well you work with machines.
What is the one thing you can’t imagine running a business without it today? That’s how AI will feel in 3–5 years. We’ll wonder how we lived without it.
AI co-founders are coming.
Stellar companies will be led by AI-first CEOs or even AI CEOs. Already happening though.
And here’s another perspective:
Conway’s Law now applies to humans + AI.
(Quick refresher: Conway’s Law says your product mirrors how your team communicates.)
So, if your company is fragmented, your AI usage will be too. If your team treats AI like a stranger, your product will feel like one.
Alignment isn’t just human-to-human anymore. It’s human-to-AI too.
Most teams use AI like duct tape.
Shopify’s using it like oxygen.
This is the shift: from tinkering to transformation.
AI isn’t a sidekick anymore. It’s becoming part of the air we breathe at work. It’s quietly integrating into how we build, design, write, support, and ship.
You won’t always notice it. But it’ll be powering:
Yes, we’re still early.
We’ve only just caught the first glimpse of what agents can do. But mark my words: AI’s impact will exceed the Industrial Revolution’s.
I believe in the next 2–4 years, near-universal adoption is likely. Some exceptions will remain but not because they can’t adopt AI, but because they won’t. Because of inertia. Legacy habits. Old-world company charts.
Here’s where we stand, according to McKinsey (early 2025):
And the real momentum? It’s coming from builders:
Nearly 90% of new notable AI models in 2024 came from the industry, not academia. That’s up from just 60% the year before. (Source: 2025 AI Index Report)
And that’s just what’s public. Away from the public spotlight, in the top research labs and dev shops, coding agents and research copilots are already reshaping how work actually gets done.
Here’s proof that AI is already quietly transforming operations across the board:
Above, see how AI is already quietly becoming “oxygen” across functions. From customer service to inventory to hiring, AI is the silent engine behind how modern companies move.
Credit: Forbes — The graphic shows how widely spread AI is across different functions
What if most startups don’t fail because they lack product–market fit…but because they fail to evolve how they build fast enough in the AI era?
It’s not just about what you build anymore. It’s about how quickly you adapt to building with machines.
And the good news? AI power is compounding — but access is democratizing. Today, OpenAI’s pro tier is ~$200/month. Agent-style coding tools like Devin? ~$500/month. Sure, top-tier tools launch expensive. But something wild is happening underneath:
Last year’s AI superpower becomes this year’s baseline — at 1/50th the cost. Just look at DeepSeek R1, launched in early 2025. It delivered GPT-4-level performance for a fraction of the price; around 1/50th the cost. That’s not just evolution. That’s deflation of capability cost at AI speed.
AI talent — once reserved for unicorns and FAANG gets cheaper and more accessible every quarter. Which means the barrier isn’t money. It’s mindset and motion.
The teams that learn fastest, win.
AI won’t just flatten cost curves. It’ll flatten companies that don’t adapt their speed, habits, and structures to match.
So, what does all this actually look like inside a modern company? Let’s break down the key mindset shifts from Shopify’s playbook — and how you can start applying them today.
Before hiring, check if AI can already do the job.
Need more staff? Prove why AI can’t do it first. More is no longer merrier. AI forces clarity on who (or what) should do the work.
Use AI as your first instinct not your fallback before adding headcount. Before you ask for headcount, ask this: “What would this role look like if an AI agent already worked here?”
It’s not about replacing people. It’s about leveling up how we decide who (or what) should do the work.
Why it matters:
In an AI-powered world, hiring blindly is wasteful. This forces clarity, creativity, and better ops decisions before adding headcount.
At the end of every sprint, reflect:
Learning is self-driven but speed scales when shared. Progress compounds when prompt wins, failures, and tools are shared in the open. Ask your team:
We’re entering a phase where we’ll all have to re-qualify for our own jobs. Not because AI is coming for us. But because working without it won’t cut it anymore.
Offload routine decisions to AI so humans can focus on higher-order thinking.
AI is a decision compressor. It collapses routine judgment — research, comparisons, quick critiques — into seconds. This frees up your most valuable resource: human taste, clarity, and narrative thinking.
Why it matters:
The less time you spend on low-stakes calls, the more energy you have for high-impact ones. Let machines excel in well-defined, data-rich environments, while you as a human continue to excel at handling ambiguity, ethical nuances, and novel situations.
Embed decision-compression into your work using models like:
I don’t know what’s the matter with people: they don’t learn by understanding; they learn by some other way — by rote or something. Their knowledge is so fragile!
– Richard Feynman
At the start of each sprint:
Run ChatGPT or Copilot over your most recent to-do list. Highlight what it could’ve handled. That’s your first wave of automation.
Ask Yourself:
Using AI shouldn’t be a last resort — it should be muscle memory.
At Shopify, reflexive AI usage is the new baseline now. t’s no longer optional , it’s your multiplier. It means invoking AI the moment you feel uncertainty — not as a last resort, but as a first instinct.
This isn’t a tool to reach for. It’s a habit to build.
Why it matters:
Prompt craft compounds. Every refined prompt today becomes tomorrow’s reusable leverage.
Every design, doc, or commit should link to the prompt(s) that generated the supporting work. This builds:
At sprint retro:
Log both successful and failed prompts with brief outcome notes. This turns personal tinkering into institutional memory.
Start Small:
Before opening your design/code editor, write one prompt describing the block of work you’re about to tackle. See what AI gives you back. Ask your team to show you the prompt that got you here.
In the AI era, working alone is optional. The habit of reaching for a co-pilot not after you’re stuck, but before you begin is the real edge.
Use AI to generate fast, divergent ideas before committing to build.
“AI must be part of your GSD Prototype phase.”
My favorite. Build with AI as a core mindset. You’ll explore more options, faster. It saves energy and makes you wildly more creative. Sprint smarter, not just faster.
Prototypes aren’t for shipping — they’re for learning fast. AI collapses a week of exploration into an hour of iteration. Your team’s learning velocity becomes a function of AI fluency ⇒ f(AI skill).
Why it matters:
It shrinks learning cycles from days to hours. Instead of betting everything on one idea, you can explore 3–5 directions before you’d normally even ship one. The first system design is rarely the best, so as you learn, your whole approach might need to shift.
That’s exactly where Conway’s Law kicks in: Your organization’s flexibility directly impacts how well you can adapt and build the right thing.
Start every sprint with a 1-Hour Agent Sandbox:
The goal isn’t polish. It’s proof or disproof.
No matter where I’ve worked, I’ve always carved out this type of exploration blocks. Dedicate the first 10–15% of the sprint to “Fail Fast with Agents.” Focus this time on:
This compresses your build–measure–learn loop into near real-time.
Ask Yourself:
Speed without learning is waste. But now you have a teammate who can generate, test, and iterate on demand. Use it or move slower than those who will.
Empower your best people with the best tools, then spread what works.
“Tools become 10× themselves… get 100× the work done.”
We’ve seen 10× individuals drive outlier outcomes. Now give those same people 10× tools and you unlock second-order compounding:
10× talent × 10× tools = 100× impact
This is how leverage actually works in the AI era.
Every team has that one person who consistently unlocks momentum. Find them. Then arm them fully with bespoke AI tools and workflows.
Once they’re moving 100× faster, spread what works:
Ask Yourself:
Compounding isn’t just financial anymore. It’s operational. And the sooner you build around it, the faster everything scales.
Prompting well is becoming a core skill — build systems to improve it.
Don’t wait for a formal training course — just start experimenting. There’s no such thing as a perfect prompt. Treat AI like a whiteboard: use it messily, quickly, and often. The learning curve only looks steep until curiosity kicks in.
The best teams learn together, out loud.
If your learning velocity isn’t compounding, you’re falling behind. That’s the new standard.
At Shopify, prompt craft isn’t a side skill, it’s part of the job. Employees are now reviewed on:
In other words, AI fluency is becoming a career moat.
Every week, track:
Review it every Friday.
Ask Yourself:
At stand-up, take 3 minutes for one teammate to demo a refined prompt. Give quick, structured feedback:
This builds a shared language for AI craftsmanship.
PMs and devs should regularly review key prompts to:
Ask Yourself:
The best teams won’t just ship faster. They’ll think better with machines — and that starts with learning how to talk to them well.
Like most things in life, you don’t leap from zero to one overnight.
Lütke’s framing is bold and future-facing — and I’m here for it. But let’s not skip the messy middle. There are real-world constraints, frictions, and blind spots that deserve attention too.
Some grounding questions I’d ask:
Not everyone has the tools, mindset, or context to apply AI meaningfully in their work. And that’s before we even get into the ethical fog: data risks, security gaps, and privacy blind spots that quietly shape outcomes.
2. Are we enabling people to build internal tools, or just asking them to use AI tools built elsewhere?
There’s a world of difference between using AI and building with AI. Teams that can tweak, prototype, or shape internal tools — even with no-code — are more likely to integrate AI meaningfully, not just superficially.
3. Do we have the right balance between autonomy and alignment in AI adoption?
Lütke talks about giving people agency. But without clarity on when and how to use AI, autonomy can lead to fragmentation. Are teams experimenting freely, or just flailing around?
If AI’s supposed to empower everyone, why does it still take a few to make it useful for the rest?
Hint: mindset?
When AI handles the repeatable (which we call commodity layer of work) , your job is to focus on the irreplaceable. This then gives you the chance to gain back the scarcest resources: time, focus, and judgment.
That’s what you reinvest into taste, narrative, and staying close to the customer — the levers no model can (yet) automate with intent.
In AI-native companies, the question isn’t how to adopt AI. That part is inevitable.
The real question is: How do we stay ahead — by using AI not just with speed, but with clarity, creativity, and conviction?
Mastering prompt craft isn’t a side skill anymore. It’s how we align AI output with real business needs and human intent. This moment isn’t just technical. It’s strategic.
AI literacy + creativity = the most valuable combo in a generation.
This brings me to the real edge: meta-decisions.
Decisions about how decisions are made. Not just what you decide — but how you decide.
That’s what keeps teams aligned, consistent, and moving fast in the right direction. When AI becomes second nature, it clears the runway for the hard calls — the ambiguous, high-stakes trade-offs only humans can truly synthesize.
And that’s the point.
The future — even today — belongs to the AI-native builder.
So why wait? Act like AI is already on your team. Not just a tool, but a thinking machine — ready, willing, and surprisingly helpful. Whatever waters you’re swimming in.
The choice is yours:
Treat AI like your teammate. Or wait until the market reminds you.
If you made it this far, you’re exactly the kind of person I write for. Drop me a line, leave a 🏄🏻♀️, or just hit the clap button. To let me know you’re out there. Thanks for being here.
P.S. I’m starting a newsletter and I’m excited about it. Curious? [Here is the edge.]
From code to character: What a $500 Robot Gets Right About AI// A $500 robot just did what billion-dollar AI demos couldn’t. Reachy Mini shows us why: One head tilt. One blink. One giant leap for lova...
The viral chart nearly everyone read wrong.// 3–4 min read · A single chart convinced the internet that students run AI. One deeper look that rewrites the story.First appeared in Essentialist Edge · A...
Begin Before Anyone Believes — Be It Deep Tech or MountaineeringFrom death zones to a YC-backed $21M orbital bet: decoding conviction in deep tech.Disclaimer: I’m not a space expert. I didn’t grow up...
