Sixty Years of Design: costs and trade-offs

The poster cost €8,400 to make in 1966. A hundred and twenty hours of studio time: a photographer booked for a full week, a typesetter given a two-day lead, a mechanical paste-up delivered by hand to the printer, a press check attended in person. The work was physical, cumulative, and largely irreversible. If the typesetter set the wrong weight, you waited two more days.

Today, the same poster takes fourteen hours and costs €1,650. Most designers accept this reduction as a natural condition of the trade — like faster internet or cheaper compute. But the numbers describe something more structural than a price drop. They describe a systematic transfer of skill: from people to tools, from trained bodies to trained models, from expertise earned over years to operations completed in seconds.

The compression happened in discrete steps, not a smooth decline. Desktop publishing in 1985 eliminated the typesetter and the mechanical paste-up — two careers, removed in one decade. Digital photography and online stock libraries in the early 2000s began to erode the commissioned shoot. Cloud design tools in the 2010s removed the courier, the two-day revision cycle, and the printing trade's monopoly on colour proofing. Generative AI since 2022 has begun to eliminate the image itself as something that requires a photographer, a studio, a lighting rig, or a human eye behind a lens.

Each wave took out a category of skilled labour. Not the designer — not yet — but everyone in the designer's supply chain. The photographic retoucher. The hand-lettering artist. The mechanical artist. The reprographics specialist. The studio messenger. All of these roles existed, were well-paid, and required years of training. None of them exist in any meaningful volume today.

The six stages of a commercial poster — brief, ideation, image-making, layout, review, production — still exist in 2026 as they did in 1966. What has changed is which parts of each stage require trained human perception, and which have been absorbed by software or model. The interactive table below maps both eras in full: same stages, same question, sixty years apart.

Click any task card to open its detail — what the work involved, how it was done, and what changed.

The brief. The final selection. The signature. Across every wave of compression — DTP, digital photography, cloud tools, generative AI — these three moments remained human. The conversation in which a client's actual need is distinguished from their stated need. The moment of committing to one direction from many candidates. And the formal liability that attaches to a person when the work goes out into the world.

These are not permanent because they are artisanal or romantic. They are permanent because they require context that cannot yet be described precisely enough to hand to software. The client's history. The cultural moment. The designer's formed sense of what the category is tired of, and what it has never seen. These forms of context are tacit, accumulated over time, and not yet transferable to a model.

Typesetting: commodified by 1990. Photography: partially commodified by 2010, substantially by 2022. Retouching: commodified by 2018. Format adaptation: commodified by 2023. The pattern is entirely consistent across sixty years of compression: any task that could be described precisely enough to write software for became, in time, free or nearly free.

The uncomfortable implication is that most creative work can be described precisely enough to write software for. The exception is not execution skill — it is aesthetic judgment under ambiguity: the capacity to decide what the work should feel like before you can say why.

The tasks that remain human in 2026 — strategic brief interpretation, concept direction, final sign-off — are expensive because they require context software doesn't have. But they are not beyond the reach of AI. They are simply not yet within it.

What tends toward commodification from here: audience analysis, competitive scanning, brand voice generation, concept generation at volume, and — more slowly — the judgment of which concept to pursue. Each of these is already partially automated. The question is not whether they will be automated further, but at what quality threshold the automation becomes good enough to trust.

Agency remains where judgment resists specification. The designer who can say "this one, not that one" — and be right often enough to be worth paying — is not easily replaced. But the surface area of that judgment is contracting with each model release.

The shift is not from designers to no designers. It is from a trade defined by mastery of execution to a trade defined by mastery of selection. The irreducible advantage is taste operating at the edge of what a brief can fully specify: the capacity to make a choice that justifies itself in the final object, not in the reasoning behind it. That is the residue. It is real. It is worth protecting. But it is smaller than it used to be, and getting smaller.

Figures are illustrative estimates for a single A2 commercial poster, normalised to 2026 euros. Sources: AIGA / D&AD historical rate references, period studio accounts, contemporary SaaS pricing, current generative-AI per-image costs.