Ten Drafts. Now What?

Thesis

A design system consumed by generators changes more than the tooling landscape. It changes everyday life in product teams. Three shifts are becoming visible in daily practice: design reviews turn into evaluations, the classic design-engineering handoff disappears, and the scarce resource in the team is no longer production but judgment.

The first part of this series describes the structural movement on the brand and system level. This second part translates it into the roles, rituals and development paths of a product team – as a diagnosis of what is currently observable, and as a sketch of the development fields that follow from it.

Three Scenes from Everyday Product Work

Scene one. A product designer looks at five variants of an onboarding flow, produced by a generator in eleven minutes. Which one goes to implementation? The answer “I have a good feeling about this one” no longer holds up: it can neither be documented nor repeated, and others cannot retrace it. The decision needs criteria the team can verify.

Scene two. A product manager brings a generated prototype into sprint planning. The prototype works, matches the brand system at first glance, is buildable. The engineer asks: “So is this the design decision now?” Nobody in the room can state precisely who made which decision, and on what basis.

Scene three. A design lead sits in a quarterly meeting facing the question of why the design team still needs three people at current staffing. The justification of recent years – “we produce the interfaces” – no longer convinces: a PM and an engineer with a generative tool stack produce an interface faster. A different justification would stand: “The team owns the judgment work and the system governance.” Except that in most organizations this justification exists nowhere in documented form.

The three scenes are not dystopian. In product teams with the corresponding tool stack they are already routine. What remains open is only how teams adapt their roles and rituals to it – and where their juniors will learn to judge in the future.

What Shifts in Three Roles

Product Designer

The core competence moves from craft depth to decision architecture – a relocation of leverage, no devaluation of the craft; the craft remains the very ground from which judgments become possible. Where a finished artifact used to be the result at the end of the day, it is now a reasoned decision, which the generator then executes.

Three competences gain weight in the coming phase:

• Spec literacy. Formulating requirements in a form that generators can consume productively. This differs from classic user stories: structured context, clear acceptance criteria, explicitly named edge cases and counter-examples. Related to what Amy Hoy and others describe as Sales Safari – except that the safari result here does not end up in ad copy but in generator context.
• Evaluation. Checking generated output against problem intent, brand conformity and contextual fit. Judgment work that rests on criteria rather than gut feeling – and that can be trained once the criteria become explicit.
• Position visibility. Publicly justifying one’s own decisions – as professional practice, not as a personal-branding reflex. The Nielsen Norman Group has spoken of thought leadership as practice for years; here the term takes on operational meaning.

Design Lead and Design Ops

The role moves from team enabler to system curator. Three fields of work step into the foreground:

• Owning the runtime setup. Structuring the design system not as documentation to maintain, but so that generators consume it correctly. Tokens with semantics, rules with counter-examples, templates with context variants.
• Institutionalizing evaluation. Criteria catalogs for generative output, review rituals with checkpoints, documentation of evaluation decisions. Without this infrastructure, output quality dilutes below the threshold of perception until a rebranding agency notices the damage.
• Defending the budget. The statement “AI does that anyway” has reached many boardrooms in 2026. Design leads need resilient arguments for where human judgment makes the measurable difference – not rhetorically, but empirically.

PM and Engineering

The adjacent roles become more design-capable, not less design-dependent. A product manager who has three flow variants generated is no exception in 2026. Neither is an engineer who builds a feature in brand conformity with an open component library and a token file.

That is no threat to design, provided design adapts its role consistently. It becomes a threat when design clings to the old role definition: “We produce the artifacts.” The future of the role does not lie in what another function can also produce with a tool.

The classic design review doesn’t scale with generative speed. Evaluation is the new review.

Four Operational Shifts

What concretely changes in everyday team life comes down to four movements:

Research stays – and gains value. Generators produce many variants; what they cannot produce is knowledge about what people in a specific context actually need. Qualitative research, structured interviews, context studies – these inputs remain human. Teams that invest in research competence now buy themselves leverage in a market that devalues production and revalues problem understanding.

Design reviews become evaluation sessions. The classic review discusses work that designers produced. The evaluation session assesses output produced by generators – against criteria, against intent, against context. A different type of meeting: shorter, with explicit criteria, built for decisions. The best teams develop criteria catalogs; the less good ones keep debating gut feeling.

Handoffs disappear. Where specification, prototype and implementation converge – and in the current generation of generator products they long since do – the classic design-engineering handoff falls away. Vendors now build the one-step transfer from prototype to implementation as an explicit product promise; the analysis is in the first part of this series. With the handoff, the meeting landscape, the tool stack and the role definition change too. Teams that still ritualize Figma-to-Storybook handovers in 2027 burn time without return.

Quality assurance moves into the infrastructure. Brand conformity, accessibility, consistency across features – these dimensions are increasingly solved structurally, through the runtime system, not through one-off individual reviews. That is ops work. Good infrastructure makes a small team act bigger.

The Unsolved Problem: Junior Education

The classic learning path in design worked through volume: build two hundred buttons, and you understand what a button is. Design twenty onboarding flows, and you know which patterns carry. Volume was the teaching stage, and it lived in the mid-level market.

When the mid-level market shrinks because production moves into generators, the learning environment disappears with it. That is an education problem, not a tool problem. Three approaches are taking shape, none of them complete:

1. Pairing with a senior plus a generator. Juniors work on real projects, with a senior person as judgment coach and a generator as production tool. It costs senior time but produces real learning curves. The Harvard Business Review describes similar configurations in its fractional leadership series: senior expertise scales across several contexts when infrastructure absorbs the operational load.

2. Public portfolio work. Juniors build their own projects, document decisions publicly, train judgment on material they own. It requires self-discipline and access to feedback communities. It works for part of the cohort, for others it doesn’t.

3. Curricula focused on judgment. Universities and bootcamps that interweave generator fluency with judgment building – instead of playing tool training off against craft. In Germany the first programs of this kind are still in development; whoever builds them holds a structural head start in three years.

The junior gap concerns companies, universities and professional associations together – and none of the three will close it alone.

Production gets cheap. Judgment gets scarce. Whoever trains judgment builds design’s future.

What Organizations Should Do Now

The decisive distinction is shifting. Until now the question was: How many designers do we need? Going forward it is: How much judgment work and system governance do we need, and in what configuration?

A mid-level designer without a judgment position produces in 2027 what a PM with a generator also produces – interchangeable. A senior designer with evaluation practice and system competence produces what a PM cannot: judgments under uncertainty, brand enforcement in edge cases, system decisions with a long half-life – not interchangeable.

For smaller organizations that cannot afford a full-time head of design, the fractional configuration becomes more attractive: strategic design leadership part-time, with a runtime infrastructure that absorbs the operational share. The model only works if the system stands before the role is filled – not after.

Perspective

This piece is written from the perspective of a product design lead in the DACH region, with a focus on SMEs, digital units and mid-size organizations. The scenes and shifts are observable in teams with the corresponding tool stack and a certain willingness to adopt tools. What this perspective does not cover: heavily regulated industries with multi-year certification cycles, corporate structures with a dominant in-house culture, non-European markets.

Critical Assessment

What holds up

• The operational shifts are already observable in teams working with a generative tool stack, not speculative
• The junior problem is historically documented in comparable reproduction upheavals (desktop publishing, digital photography) – the middle layer shrinks, and the teaching stage disappears with it
• The distinction between productive and evaluative design work has methodological precursors in cognitive task analysis research (Gary Klein and others)

What needs qualification

• Tool adoption in DACH organizations runs considerably slower than in US contexts – timing judgments based on US data transfer only partially
• Whether evaluation actually becomes “the new review”, or erodes in teams into a bureaucratic variant of a review, is open. Criteria catalogs can degenerate into checkbox rituals
• “Judgment” as a core competence only remains robust if it rests on craft. Judgment without production experience quickly loses substance in practice

Discussion Questions

01 Learning paths: Which learning structure replaces mid-level production work as the teaching stage for designers in 2030 – and who is responsible for building it?

02 Evaluation criteria: Which criteria for brand conformity, contextual fit and product intent can be operationalized, and which remain a matter of judgment?

03 Budget argumentation: How do you measure the value of judgment work in a cost structure calibrated to output volumes and tool efficiency?

04 Role distribution: Does system governance belong to the design discipline, or to a new role positioned between design, engineering and product operations?

Sources

• Anthropic: Introducing Claude Design by Anthropic Labs (analyzed in the first part of this series)
• Nielsen Norman Group, Evaluating Generative AI Interfaces (ongoing series, 2024–2026), nngroup.com
• Gary Klein, Sources of Power: How People Make Decisions, MIT Press 1998 (foundation of cognitive task analysis, a methodological precursor of the production/judgment distinction)
• Harvard Business Review, The Rise of the Fractional Executive, 2025

Glossary

Spec literacy The ability to structure requirements and context so that generators can work from them productively. Covers acceptance criteria, counter-examples, explicitly named edge cases.

Evaluation (in the context of generative output) Assessing generated results against defined criteria – as distinct from the classic design review, which is calibrated to discussing human work.

Handoff The transfer of design artifacts to engineering for implementation. A classic workflow step that increasingly falls away in tools with converging spec, prototype and implementation.

Fractional executive A leader who works part-time for several organizations at once – not as a consultant, but with strategic responsibility. Established in the US since the 2010s, growing in Europe since around 2023.

---

Second part of a series on how the design profession shifts when design becomes executable. Part 1: Brand Becomes Software · Part 3: The Backlog Was Yesterday.