What Happens When Building Interfaces Becomes Almost Free? Seven bets on AI, product design and the way we will build software by 2030

When I started working in design, creating a digital product required a surprising amount of manual labour. We built layouts in Photoshop. We exported assets in several resolutions. We documented interactions that could not be shown directly. We prepared files for developers and then explained what those files were supposed to do.

A significant part of the job was not really about solving product problems. It was about overcoming the limitations of the tools... Then the tools improved.

Sketch made interface design more focused. Figma made collaboration easier. Components, auto layout, design systems and interactive prototypes reduced repetitive work. Better handoff removed some of the friction between design and development.

Each generation of tools lowered the cost of producing an interface.

AI agents may lower it again, but the next change feels different.

We are no longer talking about a better drawing tool. We are moving towards systems that can analyse research, propose product flows, generate interfaces, connect components, create prototypes, write code, test accessibility and deploy a working version.

Potentially as parts of the same workflow.

That changes more than the speed of design. It changes what becomes valuable once the production of an interface is no longer the expensive part.

Here are seven bets I would make about product design between 2027 and 2030.

1. Producing screens will no longer be the bottleneck

For a long time, an interface was expensive to produce.

Even after a team understood the problem, someone still had to turn the solution into flows, wireframes, visual designs, responsive layouts, states and specifications.

That production work consumed a large part of the process.

AI is likely to compress it dramatically.

A designer may soon be able to describe the goal, select the relevant design system and ask an agent to create several complete directions. The agent could prepare responsive states, edge cases, content variations and an interactive prototype. It may also produce a first front-end implementation.

Work that currently takes hours or days could happen in minutes.

This does not mean that design disappears.

It means that the screen itself stops being the scarce resource.

When the cost of producing something falls, value moves elsewhere. In this case, it moves towards deciding:

• whether the screen should exist at all,
• what problem it is supposed to solve,
• which behaviour it should encourage,
• which trade-offs are acceptable,
• and how it fits into the wider product.

For founders, this creates both an opportunity and a trap.

The opportunity is obvious: small teams will be able to test more ideas without hiring a large product organisation.

The trap is that producing more screens is not the same as making better product decisions.

When interface production becomes almost free, teams may create far more product than their users actually need.

2. The visual baseline will rise

A lot of software still looks amateurish not because its creators do not care, but because good visual execution requires experience.

Spacing, hierarchy, typography, responsive behaviour, accessibility and component consistency involve hundreds of small decisions. Weakness in any of them can make a product feel unreliable.

AI tools will increasingly correct these problems automatically.

They will suggest better hierarchy, identify inconsistent spacing, improve colour contrast, adapt layouts to different screen sizes and warn about accessibility issues.

Someone with limited design experience will be able to create a product that looks competent much faster than they can today.

This is good for users.

There should be fewer products with unreadable forms, broken mobile layouts and completely inconsistent navigation.

However, a higher baseline will create another problem: sameness.

If thousands of products are generated using similar models, trained on similar interfaces and guided by similar best practices, many of them will look perfectly acceptable and almost identical.

Being visually polished will no longer be enough to stand out.

The real differentiation will come from the product’s point of view:

• What does it make easier than the alternatives?
• What does it deliberately leave out?
• How does it behave differently?
• What does it understand about its users that competitors have missed?
• Does it have a recognisable character?

“Clean and modern” is already a weak positioning strategy. By 2030, it may mean almost nothing.

3. Designers will manage capabilities, not applications

The previous generation of designers was often defined by the software they knew.

First Photoshop. Then Sketch. Then Figma.

The next generation may be defined less by mastery of one application and more by the ability to combine multiple AI capabilities into a reliable process.

Imagine a workflow in which:

• one agent analyses interviews and support conversations,
• another identifies recurring user problems,
• another generates alternative product flows,
• another builds a prototype,
• another checks accessibility,
• another creates motion and 3D assets,
• and another translates the approved direction into production code.

The designer is no longer simply operating a tool.

They are coordinating a system.

This begins to resemble leading a product team. The designer provides context, defines constraints, reviews output, resolves contradictions and keeps the final experience coherent.

The quality of this process will not depend on who writes the longest or most elaborate prompt.

It will depend on who can recognise when the result is wrong.

AI can generate something that looks complete while misunderstanding the problem, missing an important edge case or producing a solution that is impossible to maintain.

The valuable skill will be orchestration: knowing which capability to use, how to brief it, how to evaluate its work and when to reject the result.

4. The gap between design and development will become much smaller

There has always been tension between designing a product and building it.

A design can look excellent in a static prototype while being expensive, fragile or unrealistic in production. A development team can implement the specification accurately while still missing the intended experience.

Traditional handoff attempts to bridge this gap, but handoff is itself evidence that design and implementation happen as separate processes.

AI-assisted building may make that separation less necessary.

Designers are already using tools that turn descriptions into interactive applications. Developers are already generating UI from prompts, screenshots and design files. These workflows are still imperfect, but they show the direction of travel.

Between 2027 and 2030, moving from an idea to a working product may become a continuous process rather than a chain of handovers.

This does not mean every designer must become a traditional front-end developer.

It does mean designers will need a better understanding of how products work beneath the interface.

They will need to know:

• how components behave,
• where data comes from,
• which states need to exist,
• how permissions affect the experience,
• what happens when a request fails,
• what makes an application maintainable,
• and what separates a convincing demo from a real product.

The names of the roles may also change.

We may see more AI Product Designers, Experience Architects, Product Orchestrators or Agentic Product Developers.

The title is not particularly important.

The important change is that designing and building will become increasingly difficult to separate.

5. Interfaces will become more fluid

Most interfaces today are still predefined.

A team designs a set of screens, states and paths. Different users may receive different permissions or recommendations, but the fundamental interface remains mostly the same.

Generative UI could change that.

Instead of selecting one fixed interface for a broad user segment, a product may assemble the experience dynamically around the person’s current goal, preferences and context.

A new user might receive more explanation and fewer decisions at once.

An experienced user might receive shortcuts, more information density and direct access to advanced actions.

Someone who becomes overwhelmed by complex interfaces could receive shorter blocks of content, clearer hierarchy and fewer competing elements.

An analytical user might see additional data, assumptions and sources.

This goes far beyond adding someone’s name to the dashboard or rearranging recommendations.

The product may generate different structures, controls and flows for different people.

That raises difficult design questions.

How much variation is useful before the product becomes inconsistent?

How does customer support help users who are seeing different interfaces?

How do teams test experiences that are assembled dynamically?

How do we make sure personalisation improves accessibility rather than reinforcing assumptions about people?

How does a user learn a product if the product keeps changing?

Designers will no longer define only what a screen looks like.

They will define the system of rules through which the screen is created.

6. The screen will stop being the main product artefact

Product design has already moved far beyond visual design, but the screen remains the artefact around which most teams organise their work.

It is what appears in the presentation.

It is what receives comments in Figma.

It is what stakeholders approve.

It is what developers are asked to build.

But many AI-native products will not be experienced mainly through a sequence of screens.

Users may interact through conversation, voice, commands, notifications, automated actions, connected devices or agents working in the background.

In these products, the most important design decisions concern behaviour rather than layout.

For example:

• What should the system know before acting?
• When should it ask the user for clarification?
• Which decisions can it make independently?
• Which actions require explicit approval?
• How should it communicate uncertainty?
• How can the user correct an incorrect assumption?
• How does the system explain what it has done?
• Can an automated action be reversed?
• When should the agent interrupt the user?
• When should it remain invisible?

These are design questions even when no traditional interface is involved.

This may be one of the largest changes in the profession.

The main output of design will not always be a collection of screens. It may be a model of the system’s behaviour, permissions, limits and decision-making rules.

7. Judgement will become more valuable than production speed

AI can generate fifty interface concepts before a human designer has finished creating the first one.

It can summarise research, compare competitors and reproduce established patterns.

But generating alternatives is not the same as choosing the right direction.

Data can reveal what users currently do, but it does not always reveal what should be built.

Best practices can reduce obvious mistakes, but a best practice applied in the wrong context is still a poor decision.

The more production work becomes automated, the more important judgement becomes.

Designers will need to understand people well enough to notice when what users request is different from what they actually need.

They will need to understand business well enough not to design an experience that is too expensive to operate or impossible to scale.

They will need to understand technology well enough to recognise the difference between a durable capability and an impressive prototype.

They will also need to understand incentives and consequences.

A system may be able to increase engagement, remove friction or persuade someone to take an action. That does not automatically mean it should.

When products can adapt themselves to individuals and act on their behalf, design decisions become questions of responsibility.

The future of design is not fewer decisions

It is tempting to believe that AI will make product design easy.

It will certainly make some parts easier.

Small teams will be able to build prototypes faster. Founders will be able to test ideas without waiting weeks for implementation. Designers will create motion, code and 3D experiences without depending on a separate specialist for every task.

The amount of product a small team can produce will increase dramatically.

But the number of difficult decisions will not decrease.

It may grow.

When generating another interface becomes almost effortless, teams will need stronger reasons for building it.

When every user can receive a different experience, teams will need clearer rules for consistency, accessibility and trust.

When agents can make decisions and perform actions, someone must define their authority and limits.

When the system can optimise user behaviour, someone must decide which forms of optimisation are acceptable.

The future of design is therefore not simply a move from manual design to automated design.

It is a move from producing interfaces to defining how products make decisions.

AI will remove a significant amount of repetitive work.

What remains will be the part that was always hardest: understanding the problem, making trade-offs and accepting responsibility for the consequences.

That is my prediction, anyway.

By 2030, the best product designers may create fewer screens than they do today.

But they will probably make more consequential decisions.

Do you agree? And which of these changes do you think will happen first?