Writing topicAI and Judgment
How AI changes taste, decision-making, creative work, and the professional value of choosing well.
Everybody's Smart
37 postsWisdom is not a job description
The comforting line is that AI can't do wisdom. But wisdom is not a job description. What survives automation is a position: carrying responsibility for choices that matter.
Stand at the two ends of a decision
Every decision has two ends — starting something and signing off on the result. AI is good at the middle. The valuable place is at the ends, a seat creative work didn't always hold.
Why the middle is exactly what AI took
The tasks that stay with people are not safe because they are mysteriously human. They are safe because no one agreed to let a model carry the blame.
Building with AI makes you an owner
Building with AI is no longer just using tools. It is creating them. That moves every expert from contributor to product owner — responsible for how the work scales and integrates.
Don't let your tools die in a closet
Most clever AI tools die in a closet, never shared. Giving your tool a path others can trust is creative work stepping into ownership it did not have before.
The Mega Drive feeling
A meme compares AI coding tools to the old Mega Drive: people rush home to build for the pleasure of it. The fun is real, but underneath it, how companies get software is shifting.
The flood is going inward, not outward
The common prediction says cheap software means a flood of new apps to buy. The real flood is going inward — internal tools that never leave the company, growing with its data and its way of
The SAP lesson
Before the 1990s, big organisations ran on custom software. SAP and others traded fit for one painful standard. AI-assisted building tips the balance back toward custom fit.
Building becomes everyone's job
When a product manager can build a tool in a day, the developer's job shifts toward coaching, review, and architecture. Like graphic design after Canva, the skill moves up, not away.
How to survive the chaos
Distributed building risks sprawl: unapproved apps wired into live data. The winners build the right tools, maintain them, and know when to stop building and buy.
The fear: everything will look the same
Everyone fears AI will make all work look the same. The evidence is real. But a second effect runs the other way: the rare new thing stands out more, and travels further than ever.
Be original in one thing, borrow the rest
You no longer have to be original in every part of the work. Be original in one part — the idea — and borrow solid competence everywhere else.
A good idea now spreads faster than ever
AI tools are recombination engines. A published idea becomes raw material for countless outputs, so its influence spreads faster than ever — months, not decades.
Originality pays, just not how you think
Copyright protects the artefact, not the idea. The new reward for originality is reputation that compounds — influence, audience, consulting. Not fair, but real.
Copying is not new — it is the engine
The cycle is old: new, copy, spread, worn out, new again. Copying was always the delivery system for originality. AI just runs that middle stretch faster.
AI copies the style, not the work
Older copying duplicated the work. AI copies the style, not the work — so what fades is the aura of the making, the sense that a way of seeing belonged to one person.
So be the one who gets copied
Perfecting a copy of the current trend will not pay; the machine wears trends out fast. Make something new. Be the one who gets copied. Own the idea, borrow the execution.
AI can be an author
Barthes said the author is dead: meaning is made by the reader, not the writer. AI fits that idea almost too well. In one plain sense, it can be an author.
But it cannot be an artist
An author produces a text. An artist is a person who lives a public life. You could copy a pop star with AI, but one thing stays missing: mortality.
Why a dying man's album hits differently
Bowie released Blackstar two days before he died. Its power comes from one fact: he was dying as he made it. No AI can stand in that place.
The thing machines don't have is stakes
The artist has stakes; the machine does not. A person who takes a hard public stance can lose something real, and the audience knows it. That changes what the work means.
Why an essay is the one thing AI can't write
An essay entangles lived experience and reflection that moves the writer. AI can write almost anything, except lived experience out in the world.
Sweden already sold you this model
An H&M executive once asked me: is H&M your go-to store? That question explains a whole Swedish model of business, and why AI tools feel so familiar.
Lagom means just enough
Lagom means just enough, measured against what the group needs. AI runs on the same idea: for about twenty euros a month, a skill that used to take years and money is open to anyone.
Everything it makes feels a bit the same
The tools make good work, but it tends to look alike. AI gives you the average of everything it learned — pleasant, reasonable, and almost never strange.
When everyone is good, good stops counting
When the basic level of good work rises, the middle gets crowded. Competent work stops setting you apart, and the range of what gets made shrinks even as each piece improves.
What is enough actually for?
Lagom was never about mediocrity. It was about enough for the group. A draft is not yet a decision — two human questions remain: what needs to happen, and are we willing to stand behind it?
The folder knows more than the room
Your organisation already holds far more than any meeting could, scattered across drives, CRM, and old research. AI can read all of it. So the source of truth moves from the room to the fold
Now the canvas arrives already full
The canvas is not dead. But it now arrives already full, drafted by AI from the folder. The room's job shifts from producing ideas to judging a finished draft.
The room's new job is to disagree well
A ready-made summary looks official, so rooms accept it too fast. The new session has to be designed for the opposite: to find what the draft missed.
The folder gets smarter each time
Each session's output now returns to the folder and feeds the next one. The work compounds. Your value moves from filling the blank space to finding what the data could not know.
When AI pre-writes the self
danah boyd calls it parasocial media: television performed on social media. AI is the next step, and it leaves one question — what part of you is left that the machine does not have?
The grief is real, and that's fine
Many of us resist AI, and the reasons are fair. But a lot of design work was routine checking, and AI does that part well now. That frees your attention for the work AI cannot do.
Four words we keep mixing up
Style, taste, vanguard, and imagination get used as if they were the same thing. Telling them apart shows exactly where AI helps and where it cannot.
Why AI runs a step behind culture
AI learns from things people have already recorded. The newest part of culture is not recorded yet. So AI always sits a step behind the new.
Wild data: where new material comes from
Wild data is the new material in culture that has not been recorded yet. It starts as first-hand experience, and it is closer to you than you think.
Designing is not prompting
Old tools let you work things out by hand. AI tools ask you to decide first. The real skill is forming a clear picture before you start.
Methodologies
5 postsSixty Years of Design: costs and trade-offs
From paste-up to prompt: what sixty years of compression tell us about where creative work still lives — and where it has already left.
Discovery as a continuous condition
The traditional discovery toolkit treated research as a phase. AI-native discovery treats it as a condition — a sensing layer that compounds. Here is what that shift looks like in practice,
The Three-Actor Canvas
The Value Proposition Canvas was built for two parties. AI-native concepts require three. Customer, Business, and Agent — each with their own profile, each with their own design problem.
The Shape of Work Has Changed
Six dimensions map what shifted in consulting, strategy, and design work — and where the value actually sits now.
Work anyway: a methodology for when the tech changes before you're done
Agile assumed the technology was stable at sprint start. That assumption is now false. Here is what replaces it.