AI doesn't replace the designer, it reveals them
Photography didn't kill painting. The personal computer didn't kill design — it transformed it, expanded it, and ultimately clarified what a trained designer could do with these tools. Every technical revolution in creative fields follows the same arc: anxiety, redistribution of tasks, then the revelation of what human expertise produces that the machine — itself produced by humans — cannot produce alone.
Generative artificial intelligence is at this crossroads. The question isn't "will it replace designers?" but "what does it reveal about what we actually do?"
What AI does well — and what it doesn't
Let's be direct: generative AI performs well on formal generation tasks. It explores variations at a speed no designer can match alone. It lowers the entry cost of prototyping. It processes considerable volumes of visual data.
But a relevant design solution isn't reducible to a generated form. It rests on a deep understanding of context: the implicit stakes of a brief, the contradictions between functionality and desirability, a market's cultural resistances, the ethical responsibility towards users. These dimensions can't be automated. They're acquired through the experience of completed projects — and certain failures — through years of dialogue with real clients and real users.
Matthieu Ferry, a clinical psychologist specialising in cognitive, emotional and behavioural processes, observes this phenomenon in other professional fields: the tool augments execution but doesn't substitute for judgment. What remains irreducibly human isn't only the capacity to evaluate a result — it's the capacity to assume and bear responsibility for one's choices. An experienced designer doesn't just know how to produce: they decide, they arbitrate, and they answer for their decisions before a client. AI generates options. The designer puts their signature on the line.
It's this intellectual expertise that AI augments without replacing.
The homogenisation trap
The debate about AI isn't only about design. Photographers, illustrators and graphic designers have been living under the same pressure for several years: clients replace commissions with subscriptions to image banks. In the era of platforms, these practices have spread among creators themselves. Today, clients — companies and public bodies alike — entrust the production of their visuals to generative tools, convinced that a monthly subscription replaces a creative service. The result is visible — risible? — to the naked eye: a proliferation of images that look alike, smooth, generic, technically acceptable but semantically hollow.
This is no accident. Mediocritas, in Latin, means the middle, the average. Generative AI structurally converges towards the statistical average of its training data. What Alain Deneault calls "mediocracy" — the reign of the middling — finds in generative tools its most efficient instrument.
But what can AI actually do in the hands of a trained designer? When a design professional takes hold of generative tools, something else happens: AI doesn't replace judgment, it accelerates it. It opens directions that a trained eye knows how to select, combine and redirect. It's this tandem — generative tool steered by cultivated human expertise — that produces distinctive results. AI alone converges towards the average. The designer alone is constrained by time. Together, they can go further, provided the designer remains the pilot of meaning and style. A photographer, an illustrator, a designer who takes hold of AI imprints their visual culture, their singular gaze. That's what distinguishes an image generated by an experienced creator from an anonymous prompt: intention, coherence, voice.
The phenomenon of levelling down has a name in brand strategy — loss of distinctiveness — and it predates AI. It reveals a structural tension: every brand wants both to be recognised — social desirability, "doing what others do" — and to stand out. Professional creativity resolves this tension through what might be called proximal distance: close enough to the codes to be intelligible, different enough to be memorable. Generative AI, by construction, produces proximity without distance.
Reduced budgets, rushed briefs, decisions made without a designer at the table: formal impoverishment existed well before generative tools. AI didn't create this problem. It amplified and accelerated it by giving it the appearance of rational economy.
Two discourses must be distinguished here — they have nothing to do with each other. On one side, the real potential of AI as an innovation tool in the hands of competent designers. On the other, the marketing discourse of generative tool publishers directed at designers' clients: a discourse that sells subscriptions by claiming that creation reduces to generation, that expertise can be replaced by a prompt, that quality is obtained without investment.
Yet, as Brigitte Borja de Mozota shows in her work on design management, a distinction must be drawn between visible design — the deliverable, the form, the identity — and invisible design: strategic competence, project culture, the successive decisions that precede and structure every result. Visible design is copiable. Invisible design is inimitable. It's precisely this submerged process that generative tools short-circuit, and that clients underestimate when they confuse a creative service with a subscription.
This commercial discourse produces predictable effects: organisations that replace a professional relationship with a subscription obtain visuals that look like their competitors'. A few months later, they wonder why their visibility has declined.
Because design isn't a cost. It's an investment — and the intellectual property that results from it is a tangible asset of the organisation, recordable as such in its economic value. An original visual identity, distinctive packaging, a patented interface: these are assets that can be valued, protected and transferred. The royalties paid to a designer aren't additional fees: they're the mechanism by which the organisation acquires full ownership of a creative asset.
Behind the pressure on prices lies a well-documented psychological mechanism: loss aversion pushes decision-makers to minimise visible expenditure rather than optimise return. They want a guarantee of success — which doesn't exist. This is precisely what royalties resolve: by linking the designer's remuneration to the performance of their creation, they transform a cost into a shared investment. The designer's interests align with those of the organisation, over time, not in a one-shot logic. From the designer's side, the reasoning is symmetrical: refusing the royalty model structurally selects clients who think in one-off terms — precarious relationships where each project is an isolated bet rather than a shared construction.
A designer who can explain this to their client is no longer negotiating a rate — they're making the case for an investment. It's a radically different posture, and an infinitely more solid one.
The return to human expertise won't be an act of militancy. It will be a communication necessity.
What training gives — and AI cannot simulate
Making a human designer a creator of quality visual arts isn't a question of access to tools. It's a question of developing critical thinking, understanding the history of context, the history of techniques, art history, the history of forms, the history of uses — and of the people for whom the designer will work.
A trained designer doesn't generate images or three-dimensional forms: they make choices. They know why the Bauhaus or the Modern Arts Movement broke with ornament, why the Polish poster of the 1960s spoke through metaphor, why a typeface creates trust or distance. This culture isn't an academic luxury. It's the foundation of creative judgment — the capacity to produce a solution that makes sense in a specific context, for a specific audience, at a specific moment.
This is the most original response that design training offers to AI. Rootedness in the history of forms doesn't produce imitation — it produces positioning. The trained designer doesn't reproduce the Bauhaus: they know where they stand in relation to the Bauhaus, and it's this awareness of their own position that makes their proposal singular. Continuity comes from depth; distinction comes from perspective.
AI trained on the past can reproduce forms. It cannot decide on their relevance. That expertise belongs to the designer.
To this culture is added a competence that is rarely discussed yet lies at the heart of the creative process: mastery of time. Not speed — AI excels at that — but the three temporal scales the designer inhabits and that the generative tool ignores entirely.
First, the long time: that of training, of accumulating references, of the progressive sedimentation of a perspective. One doesn't become a designer in a few weeks of prompts. This foundation is built over years, and it's precisely what makes the judgment of an experienced professional irreplaceable.
Then the medium time: that of the project, of immersion in the client's context, of the progressive understanding of a brief's implicit stakes. This time isn't a loss — it's the condition of relevance. A designer who takes time to observe, to question, to doubt before producing delivers a solution grounded in a reality that AI cannot perceive.
Finally, the short time: that of creative incubation. Cognitive neuroscience has documented this phenomenon: diffuse thinking, that mental state in which the brain continues working in the background after a period of concentrated effort. Letting a problem rest, stepping away, returning to it: this isn't a loss of productivity — it's an essential cognitive mechanism. Experienced creatives know it as productive letting-go, and they use it consciously. AI generates in milliseconds. It doesn't decant.
These three timescales constitute a professional competence in their own right and a value argument that designers can explicitly assert when facing clients who are dazzled by the promise of speed from generative tools.
Design as dialogue and shared transformation
Ask ten designers the same question and you'll get ten different answers. All potentially valid. None superior to the others before a choice is made.
That choice cannot be delegated to an algorithm. It rests on the meeting of two complementary forms of expertise: that of the designer, who commands forms, uses and technical constraints, and that of the client, who knows their deep needs, their organisational culture, their users, their real constraints. Neither holds the right answer alone. It's their dialogue — and the design process — that brings it to light.
This dialogue isn't a precondition for design — it is design. A misunderstood brief produces a technically impeccable solution that is strategically useless. A client who doesn't engage in the creative relationship receives a deliverable, not a transformation. The quality of the exchange between designer and client directly conditions the quality of the result, whatever tools are used.
Well-conducted design is a shared transformation. The designer doesn't emerge unchanged from a successful project: they've learned something from the client, from their sector, from their constraints. Nor does the client: they've sharpened their understanding of their own needs, their identity, of what they want to project. This co-construction doesn't happen with a tool. It happens between two engaged intelligences committed to the same project.
In this sense, AI also operates a transformation on the designer who uses it. Formulating a precise prompt forces the designer to make explicit what was previously intuitive. Selecting among hundreds of generated variations sharpens the critical eye. Rejecting what doesn't work clarifies what one is really looking for. Collaboration with the generative tool isn't neutral for the one who conducts it: it reveals the designer's own criteria, their own standards, their own voice. It's a form of reverse learning — provided, again, that the designer stays in command.
What this concretely changes for your pricing
This is where the thinking becomes operational.
If AI reduces the time spent on certain phases — formal exploration, generating variants, rapid prototyping — it doesn't reduce the value of the service. It shifts the designer's time towards what has always been most valuable: defining the problem, arbitrating between constraints, validating choices, bearing responsibility for the final deliverable.
Billable time changes in nature, not in value. Billing by the hour for tasks executed by a machine makes no sense in design. Billing for the value produced by expertise of judgment makes all the more sense.
The pricing pressure is real. Some clients are beginning to compare a designer's fees with the cost of a subscription to a generative tool. That's confusing a prototyping tool with a complete intellectual service. The response to dumping isn't resignation — it must be mastery of the value argument.
A few arguments every designer can use when facing a client dazzled by the promise of generative tools. An AI subscription generates forms; a design service produces an asset — intellectual property that is transferable, defensible, and recordable on the balance sheet. The cost of bad design is measured in loss of distinctiveness, brand confusion and redesign costs: it is systematically higher than the cost of good design from the outset. The brief itself has value: the designer who helps their client articulate their problem produces something that the generative tool cannot. Finally, a professional designer engages their responsibility; a generative tool answers to no one. These arguments deserve to be developed — we'll devote an upcoming Coblog article to them.
The guardrails that don't move
Using AI in a design practice raises questions that the profession must address collectively.
The copyright question is serious. Generative AIs have been trained on corpora of existing works, often without authorisation or remuneration of their authors. The designer who uses these tools bears a responsibility: ensuring that their deliverables don't constitute inadvertent plagiarism. Alliance France Design and authors' societies are advocating for a mechanism to remunerate creators whose works are used to train models — a form of neighbouring rights adapted to the AI context. This approach deserves to be pursued at the legislative level, in France and across Europe. And one can expect measures to protect the original creation of authors in the months ahead, particularly given that there are financial means capable of addressing the precarity of creators and sustaining the conditions of existence for practitioners of a structurally uneven profession.
Transparency towards the client remains a fundamental principle. If AI intervenes in the creative process in a way that affects the nature of the service, explaining it isn't a weakness — it's the condition of professional trust.
Responsibility, finally, cannot be delegated to the machine. If an AI-assisted creation proves problematic — inadvertent discrimination, ambiguous messaging, accessibility failure — it's the designer who answers for it. Always.
What cannot be automated
AI compels us to a useful exercise: identifying what genuinely constitutes our value. Not out of defensive reflex but out of professional clarity.
Empathy with users, reading a cultural context, anticipating real uses, maintaining consistency of approach over the duration of a project — all of this remains the domain of human expertise. These are precisely the dimensions that the most demanding clients know how to recognise and remunerate.
To this must be added a competence that design education develops, yet professional frameworks still struggle to name clearly: the capacity for affective attunement with one's professional environment. The designer is a sensitive being — and that is a professional quality in its own right, not a liability to be compensated. This sensitivity irrigates the entire relational dimension of the profession: perceiving what a client feels without their having articulated it, detecting the tension between what a brief says and what it conceals, understanding the audiences the work addresses — not abstractly but in resonance with them. The designer always works within a triad: themselves, their client, and their client's audience. It's this triple relationship of listening, adjustment and translation that no generative tool can simulate. These competences are cultivated, refined through experience, and constitute a decisive part of a designer's value.
Because design, whatever its discipline, seeks to touch. To touch through the eye with image, message, typography. To touch through experience with interface, journey, service. To touch with the hands through object, material, space. This ambition of contact runs through everything we do. One might ask whether an image generated without conscious emotional intention can produce the same effect as a creation carried by the lived experience of a designer who has perceived — and decided, from their own sensibility — what it should provoke.
Photography redefined painting. The personal computer redefined design. AI in turn redefines the contours of the profession: it shifts certain technical tasks towards automation and reveals, by contrast, the depth of the intellectual expertise that remains irreplaceable.
The designers who will navigate this transformation with clarity are those who have learned to articulate their value — not just produce it.
François Caspar is the founder of Copryce and co-initiator of the Alliance France Design code of ethics. Matthieu Ferry is a clinical psychologist specialising in cognitive, emotional and behavioural processes, and founder of intelligences-plurielles.com and ia-et-psychotherapie.com.
The right clients exist, here's how to recognise them
Not all clients are alike when it comes to AI. Some have understood that visual singularity is a competitive advantage. Others chase the lowest price and will come to regret it. Before you sign, watch for these signals.
They understand the value of human design if…
They talk about their identity, their values, what distinguishes them from competitors — not just their budget.
They cite precise visual references, with words for what those references make them feel.
They accept a scoping phase before any production and regard it as useful, not a waste of time.
They ask questions about your approach, not just your deadlines.
They've already worked with a designer and can speak about it concretely.
They spontaneously distinguish between "generated image" and "design creation" without needing it explained.
They understand that revising a direction isn't a sign of poor work on your part but of an insufficiently defined brief.
They talk about their users, their customers, their constituents — not just themselves.
Be wary if…
Their first question is “how much does it cost?” before they’ve even described their need.
They show you an AI-generated image and say “I want something like this, but cheaper.”
They can’t explain what distinguishes their brand or organisation from its competitors.
They treat the brief as a formality to be dispatched.
They mention a “small budget” for a “big project.”
