We no longer take every product from zero to one hundred
The honest answer is that many clients need us less for the first 70 per cent than they once did.
AI can structure ideas, propose user flows, design interfaces, generate variants and build working prototypes. It is quick, patient and never offended when you ask for the button to move left for the fourteenth time. This is excellent. It opens up a part of product development that used to be slow and expensive.
Our task is therefore shifting. We no longer take products only from zero to one hundred. Increasingly, we help move them from 70 to 120 per cent.
The 120 per cent is obviously not a mathematical measure. It describes the distance between “it sort of works” and “it is clear, distinctive, ready for the market and capable of surviving reality”. A prototype may demonstrate a good flow. It does not automatically reveal whether the product solves the right problem, whether the business model fits, or whether users recognise the value that matters.
A prototype is an answer to a question that may not have been tested
A prototype allows us to become specific much faster. That is a genuine improvement. Instead of discussing abstract requirements, we can see decisions. We can also see which questions have not been asked.
We notice when a team is already debating the colour of the curtains before deciding where the front door belongs. We notice when five people are nodding politely while imagining five different products. And we can usually tell whether a feature is genuinely useful or simply exists because it looks impressive in the prototype.
This is a large part of our work today: helping the client uncover the core. The goal is not to design as much as possible. It is to make the right decision visible.
We turn that into practical product strategy, research and testing. Which problem comes first? Which assumption needs testing with users? What belongs in the next release? A prototype makes those questions tangible. Research helps us understand whether the answer holds up beyond the team that created it.
Experience from roughly 900 projects helps. Not because we have seen everything, but because certain patterns become visible earlier. We recognise the dependency that will become expensive, the team conflict that has been transferred into the interface, and the small ambiguity that will later travel through product, engineering, marketing and support.
A company with an existing prototype may therefore need no complete restart. A UX audit can show which parts already work and where the missing 30 per cent separates a demo from a product.
Designers are no longer the only authority on design
We should be honest with ourselves. AI can already complete a substantial share of conventional design work faster. My personal estimate is that it is faster, and sometimes better, in around 60 per cent of cases. That figure does not come from a study. It is an observation from our daily work.
In a few years, the number may reach 80 or 90 per cent. A designer waiting for AI to disappear may as well wait for the fax machine to return.
This does not diminish design. It forces us to define our contribution more honestly. A clean interface, a sensible component system or a set of plausible options is no longer automatically the part for which a client needs an agency. That work will remain important, but it will be produced faster and more often with machines.
Our job lies in the remaining percentage. Unfortunately, that does not mean adjusting the final ten pixels.
The final ten per cent is not leftover work
The final ten per cent consists of decisions that cannot reliably be derived from the average of previous solutions.
What should this product deliberately leave out? Which rule should we break so the brand becomes recognisable? Where might a flow need slightly less convenience and more trust? Which apparently unreasonable idea creates the memory the product currently lacks?
AI is exceptionally good at producing what is probable. A market-ready product sometimes needs something unexpected: an edge, a pause, an unusual phrase or a decision that does not immediately please everyone. This is not decoration or designer vanity. People do more than operate products. They sense whether something was made for them or generated from the average of everything else.
Those edges still have to work. Surprise without understanding is merely confusion. Humanity without technical precision is carelessness. The task is to find the right deviation and develop it until it can survive in the market.
That is why branding belongs in this discussion. If AI produces five plausible interfaces, you may not need a sixth. You may need an identity people can recognise. We work on visual language and tone of voice, then bring them into the product. “Looks professional” becomes an experience that actually belongs to this company.
Market readiness is larger than the interface
A market-ready product connects several layers. The value proposition must be visible. Critical journeys must work. Brand and interface need to belong together. The team must understand what gets built first. Technology, content and operations cannot remain afterthoughts.
In UX/UI web design, we therefore develop the information architecture, user journeys and interactions rather than simply polishing the surface. We address states that demos often leave out: what happens with no content, after an error or on a small screen? A design system keeps those decisions consistent as the product grows.
Accessibility is part of market readiness too. An accessibility audit helps uncover barriers in design, content and technology and prioritise the next changes. A beautiful prototype has limited value if some people cannot use it properly.
Complex products make this especially clear. Our work with SIXT, for example, extends far beyond individual screens. It includes an international design system, booking flows, accounts, communications and landing pages for many markets. Projects of that scale teach you that good design does not end at the edge of a Figma frame.
AI can support all of these areas. It can accelerate variants, prepare content, identify patterns and build prototypes. It cannot independently decide which conflict a company is willing to face or which promise it can keep for years. Those decisions still require people who accept responsibility.
Three knights at one table
For me, the new model is therefore not a competition between client, agency and AI. It is a collaboration with distinct roles.
The client understands the company, the market and its constraints. Silberpuls brings project experience, an outside perspective and the right uncomfortable questions. AI supplies speed, alternatives and a new means of production.
Client, Silberpuls and AI: the three knights of the round table. Historically doubtful, perhaps, but a surprisingly useful project model.
If you already have a prototype, you do not need to explain why it is not perfect. Bring us the 70 per cent. Together, we can find the 30 per cent that is truly missing and the ten per cent that makes the product human, surprising and ready for the market.
