Most of the hype cycle around AI is framed like a productivity miracle: fewer boring tasks, more time for the good stuff, the future arriving politely at your doorstep.
In practice, it’s been more like this:
Everyone is expected to produce more.
Everything is expected to ship faster.
Nobody is allowed to admit they feel cooked.
And if you’re tired, you’re not tired because the expectations are absurd. You’re tired because you’re not “leveraging the tools.”
That’s the core of AI burnout.
The tool itself isn’t the whole story. Obviously.
The story is what happens after the tool shows up:
“We can do this in half the time now.”
“We can generate ten options instead of one.”
“We can cover twice as many channels.”
“We can crank out a first draft instantly.”
None of that is framed as optional. It becomes the new minimum.
You don’t get time back. You get more work. Because of course you do. The prize for efficiency is more work, forever, until you die or get promoted into a job that is just meetings about work.
And it’s not even better work. It’s work that looks productive on a dashboard. It looks like progress in a status meeting. It looks like “momentum”.
It also feels like running on a treadmill that speeds up every time you manage not to fall off.
Once a tool becomes fashionable, the point stops being “does this help?” and becomes “are you using it?” Because nothing says “leadership” like turning a tool into a personality.
That’s how you end up with:
AI-written meeting notes nobody reads
AI-written performance feedback nobody believes
AI-written product updates that say nothing in more words
AI-written marketing copy that sounds like it was trained on a hundred thousand polite lies
It’s not that humans suddenly became incapable of writing. It’s that writing got rebranded as wasteful. Thinking too, while we’re here.
A lot of teams are now treating human thought like a defect. If you stop typing, someone decides you’re “blocked” and tries to fix you with a meeting.
That’s the burn.
The most exhausting thing about the current AI wave is that it attacks the last scraps of downtime in knowledge work. Not the meetings though. Meetings are forever.
The expectation is not just “do your job”. Now it’s:
Also publish your thinking
Also keep your personal brand active
Also stay current on tooling
Also repackage your work into more formats
Also be available to respond faster
AI doesn’t cause this pressure. It amplifies it.
Because now there’s always a way to generate another draft, another caption, another summary, another set of talking points.
There’s always more you could produce.
And when “could” becomes “should”, burnout is not a risk. It’s the plan.
I don’t think all uses of AI are equally gross. Some are just boring. Some are actively corrosive. Using a tool to help with admin work, or to break down a messy problem, or to get unstuck on a blank page can be legitimately useful.
But using AI to generate art and music crosses a line for me. Not because I’m scared of new tools, but because art isn’t “content”. It’s the record of a human trying to say something and learning the craft to do it.
Art is:
taste
intent
struggle
context
history
the specific shape of someone’s life
When you replace that with machine output, you’re not just saving time. You’re removing the person.
And the part that people love to ignore is consent. A huge about of generative AI is built by scraping human work and reassembling it into something that looks original enough to sell. That is not creativity. That is industrialized plagiarism with better marketing.
“It’s just a tool” is a cute argument when you’re talking about spellcheck. It’s not a cute argument when the output is pretending to be human expression.
That stuff isn’t “content”. It’s:
rehearsal and craft
collaboration
bodies in a room making choices
risk, timing, tension, silence
the specific weirdness of real humans trying to mean something
Generative AI does not “help the arts” by spitting out scripts, songs, posters, or set concepts trained on other people’s work. It helps the people who want the “vibe” of art without paying artists, without learning the craft, and without admitting they’re outsourcing the part that matters. So, like, half of LinkedIn right now.
The real harm is a combination of:
non-consensual training (scrape first, apologize never)
economic substitution (replace the paid gig with “good enough” output)
style laundering (“in the style of” but no credit, not consent, no compensation)
cultural flattening (everything converges into one beige, statistically-average slurry)
If you want the grown-up version of this conversation, start here:
And if you want a loss polite version: the business model is “we copied a library of human work, then sold the ability to imitate it”.
Sure. A paintbrush is a tool.
But a paintbrush doesn’t vacuum up millions of artists’ work, learn their styles, and then let someone type “in the style of” like it’s a free buffet.
A tool that changes the incentives and the economics is not “just” anything. If the business model requires treating human creators as training data, then the product is not the tool. The product is the theft.
One of the most impressive parts of the AI hype machine is how it makes the physical world disappear. It’s all “cloud” and “magic” and “just compute”. Which is a fun trick, because “compute” is just a euphemism for “somebody else’s power grid”.
Meanwhile, in the real world:
Data centres draw massive electricity loads (and that load has to come from somewhere)
They generate heat
Cooling that heat often means water use
Hardware gets built, shipped, and replaced on a schedule that is not exactly gentle on the planet
If you want something more grounded than “trust me bro, the vibes are bad”, here are some grown-up references:
CBC (water + opacity): AI-related data centres use vast amounts of water. But gauging how much is a murky business
CER (data centres + demand outlook): Energy demand from data centers is steadily increasing, and AI development is a significant factor
CSPS (full lifecycle overview): Unpacking the Environmental Effects of AI
Canadian Climate Institute (grid tradeoffs): Is there a smart way to integrate AI data centres into electricity grids?
CCPA (policy + backlash): AI data centres: From boom to backlash
Natural Resources Canada (efficiency guidance): New Best Practice Guide for Canadian Data Centres Now Available
Greenpeace (energy + environment + democracy): The energy and environmental impact of AI and how it undermines democracy
MIT News (useful baseline explainer): Explained: Generative AI’s environmental impact
IEA (numbers people cite): Energy demand from AI and Executive summary
Text generation can be useful. It can also be spammy. Both things can be true. But image generation has a uniquely cursed pattern:
The cost is real (power + cooling + hardware)
The output is disposable
The workflow encourages you to generate fifty versions because you can
This is not “creativity”. That’s the world’s least interesting slot machine. As yes, the scale matters. Not “one image”. The endless conveyor belt of them.
One clean reference: Making an image with generative AI uses as much energy as charging your phone.
Maybe. Sometimes. In specific use cases. But “we might optimize some things later” is not a moral exemption for “let’s triple date centre load now”.
The point is not that computers are evil. The point is that we are building a wildly energy-hungry infrastructure race, then acting confused that it has consequences.
I’m not anti-computer. I’m anti-bullshit. I’ve built my entire career on computers. This is not a Luddite thing. This is a “stop making everything worse on purpose” thing. AI is genuinely useful when it reduces repetitive labour without replacing trust, authorship, or craft.
Good uses:
Automating boring admin tasks (tagging, sorting, filing, triage)
Summarizing long threads and meeting notes (with a human sanity-check)
Drafting a first pass you’ll rewrite in your own voice
Bad uses:
Replacing human feedback (performance reviews, career conversations)
Outsourcing decisions you are accountable for
Generating art, music, scripts, posters, and “creative” output that’s trained on other people’s work
Spamming the world with infinite synthetic images because someone confused “more” with “better”
If you want to keep your sanity, you need boundaries that you do not renegotiate every time someone shows up with a new feature demo.
Here’s a policy that’s worked well for me, and might work for you.
Use it for:
outlining when you’re stuck
generating checklists for boring admin
turning scattered notes into a coherent plan
Do not use it for:
replacing your judgement
replacing your voice
replacing your relationships
If the output is meant to be believed, signed, or relied on, treat AI as unsafe by default.
Examples:
performance feedback
career conversations
policy writing
anything legal
anything where a human is on the other end and deserves a real response
This is the line for art. If the point is expression, the process is not overhead. The process is the point.
You do not have to debate your boundaries like you’re in front of a tribunal.
Try:
“That’s not a use case I’m comfortable with.”
“I’m not using AI for creative output.”
“I’ll do the writing myself. I’m fine with it taking longer.”
If someone treats that as a personality flaw, that’s not a you problem. That’s a workplace problem.
The real way out of AI burnout is not a better prompt. It’s:
realistic expectations
time to think
permission to do less
fewer performative deliverables
leadership that doesn’t confuse motion with progress
AI can be helpful. But if you’re using it to prop up impossible workloads and pretending that’s innovation, you’re not building the future. You’re just finding new ways to grind people down faster.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.