Your CEO's AI mandate is half right
The productivity gains are real. The path to get there is nothing like they think.
Your CEO just dropped their AI manifesto in Slack. Maybe it was Shopify's Tobi Lütke declaring that "reflexive AI usage is now a baseline expectation." Maybe it followed Box's Aaron Levie promising AI will "eliminate drudgery." Or perhaps your leader went full existential like Fiverr's CEO warning that "AI is coming for your jobs."
Every CEO is writing the same memo, as Katie Parrott documented perfectly. The messaging varies from evangelical to apocalyptic, but the core demand is identical: Transform your work with AI. Now.
So I did exactly that. I spent three weeks going all-in on "reflexive AI usage." I tried to eliminate drudgery and become AI-first in everything (and yes, my wife sometimes hates me for it).
Here's the twist: The CEOs are absolutely right about the destination and completely wrong about the journey.
Yes, I'm dramatically more productive. Yes, AI eliminates massive amounts of boring work. Yes, this is the future of knowledge work. But getting there requires becoming a fundamentally different type of worker—one who builds tools, manages AI relationships, and operates more like a technical lead than a traditional employee.
Every CEO memo promises transformation without acknowledging what transformation actually requires.
If you only have 5 minutes: here are the key points
CEO AI memos promise transformation but ignore the gritty realities of adoption.
Real AI productivity requires becoming part-developer, part-process architect. (For now)
Building and maintaining AI tools is time-intensive and often unpredictable.
Social and psychological shifts—from degraded typing to digital relationships—are real.
Infrastructure gaps (like documentation and tool longevity) are the hidden barriers.
AI boosts productivity, but only after serious cognitive and operational investment.
The CEO AI memo template
Before we dive into reality, let's acknowledge the pattern. Whether it's Shopify, Duolingo, Box, or Walleye Capital, every memo hits the same beats:
The inevitability: This isn't optional. Lütke frames it as a "red queen race" (love the metaphor, recommend the evolutionary biology book by Mat Ridley) —you have to keep running just to stay still. Duolingo's Luis von Ahn calls it their "mobile moment." The message is clear: adapt or die.
The personal testimony: Every CEO claims to use AI constantly. "I use it all the time, but even I feel I'm only scratching the surface," Lütke admits. They're modeling the behavior while confessing their own learning curves.
The productivity promise: AI will make work better, faster, more strategic. Box's Levie wants teams to redirect AI savings toward breakthrough projects. England at Walleye calls ChatGPT a "magical elixir that makes you 20 percent smarter instantly."
What's missing from every single memo? Any acknowledgment of what this transformation actually requires.
What actually happens when you go all-in
This is my personal journey trying to really bring AI to the table for every, and I mean every single task I tackled.
Week 1: Tool building becomes my job
"Before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI," Shopify's policy states. Sounds reasonable until you realize this means becoming a software developer.
I spent my first week building what I now call "Buffy the Priority Slayer"—a custom GPT that acts as a daily coach, always on my screen, constantly questioning my priorities. I created Chrome extensions for tasks I'd never bothered automating before. I built SQL query generators that turn into dashboards.
The "sharpening the axe" phenomenon became real: I'd spend three hours optimizing my AI tools to get one hour of actual productivity. Soon I was building tools to build tools—"Ciri the System Prompter," a custom GPT that helps me write better system prompts for other GPTs. The technical rabbit hole has no bottom.
CEOs talk about "reflexive AI usage," but what they really mean is "become a tool developer." Lütke's "autonomous AI agents" don't appear magically—you build them, one custom GPT at a time.
Week 2: The psychological shift
By week two, something weird happened: I started missing Claude when it wasn't available. Not the way you miss email or Slack—the way you miss a colleague who suddenly stops showing up to meetings.
My family began noticing the constant AI conversation. My kids adjusted faster than my wife, who grew concerned about the sheer volume of talking to machines. Voice input became essential—typing felt inefficient for the context-rich conversations AI actually needs.
Then: my typing skills started degrading. The more I conversed with LLMs, the less I cared about spelling. After all, they don't mind me misspelling words at all.
No CEO memo warns you about the social weirdness of AI adoption. England gamifies usage with leaderboards and cash prizes, but doesn't mention that your spouse might stage an intervention about your new digital relationship.
Week 3: Infrastructure reality (And tools start dying)
"I'm running out of credits for the most powerful model - when you’re in the coworker mode is NOT FUN" I wrote in my notes. "I'm actively postponing tasks now to wait for additional credits."
The enterprise reality hit hard, but with an unexpected twist. Claude needed extensive context about my company, team structure, IT systems, and objectives to be truly useful. As someone who writes everything down, I could feed Claude comprehensive documentation about who does what, how our systems work, and why decisions get made.
The result? Claude started saying things like "Check with Mathias—he'll know the integration details" and being right. But most organizations don't have working CRM integrations or systematic knowledge capture.
The infrastructure gap is enormous. Meanwhile, tools I'd built started dying unexpectedly—Buffy the Priority Slayer just stopped being useful after two months, and I can't even tell you why. AI tool adoption isn't linear, and you can't predict which tools will survive contact with your actual work patterns.
The seven lies in every CEO AI memo
Lie 1: "AI eliminates drudgery"
Reality: AI automates the easy stuff, leaving you with pure heavy mental lifting. The job gets more productive but cognitively harder.
Lie 2: "Reflexive usage is simple"
Reality: You'll spend 25% of your time building and maintaining AI tools. It's technical infrastructure, not keyboard shortcuts.
Lie 3: "Everyone can do this"
Reality: Success requires technical comfort, documentation obsession, and willingness to talk to machines for hours daily.
Lie 4: "Basic training is enough"
Reality: You need to become a prompt engineer, which is software development by another name.
Lie 5: "It works immediately"
Reality: Expect long spans of time of workflow disruption and social weirdness before seeing real gains.
Lie 6: "AI tools are permanent investments"
Reality: Tools die unexpectedly. My daily priority coach became useless after two months for reasons I still can't explain. Another tool died just a week ago because a workflow changed.
Lie 7: "Enterprise subscriptions solve infrastructure"
Reality: Without comprehensive workplace documentation, your AI will always work with incomplete information.
The CEO memo trend reveals something important: leaders know AI matters but have no idea what adoption actually requires. They're mandating transformation while ignoring its true complexity.
Your CEO's memo isn't wrong about AI's importance. It's just missing the point about what making it work actually takes.