has quickly become part of the workflow in , and web development. It can speed up research, ideation, production and development. But speed is not the same as quality.
The discussion around AI is often unnecessarily polarised: either the technology will replace creative professions, or it is an overhyped tool that only produces mediocre work. The reality is more practical. AI is already a real working tool, but its value depends heavily on how it is used.
Figma's 2025 survey looked at 2,500 designers and developers. 78% said AI improves their efficiency, and 85% said learning to work with AI will be essential to future success. Designers were more cautious than developers about quality: 54% of designers said AI improved the quality of their work, compared with 67% of developers.
AI can make it faster to produce an option. It does not automatically make that option good.
Graphic design: more possibilities in less time
One of the clearest benefits of AI in graphic design is that it lowers the time required to explore alternatives.
It can help with:
- early visual directions
- generating or editing images
- alternative copy and messaging
- quick mockups and presentations
- adapting material to several formats
- automating parts of repetitive production
Figma reported in 2024 that 59% of designers and developers in its survey were already using AI at work. Common uses included user research, image and text generation, brainstorming and ideation.
This changes the pace of exploration. A designer can inspect ten possible directions in the time it once took to build two.
The designer still has to decide whether the typography works, whether the composition has the right hierarchy, whether the visual language fits the brand, and whether the result communicates what the client actually needs.
AI can produce alternatives. Expertise is used to select, combine, reject and improve them.
UX design: an assistant, not a synthetic user
In UX, AI can be useful long before there is a finished interface.
Nielsen Norman Group conducted a study with 841 UX professionals and described four common roles for generative AI: content editor, research assistant, ideation partner and design assistant. Among respondents who used AI for work, 63% used it at least several times per week.
AI can help with:
- a first draft of an interview guide
- alternative survey questions
- summarising large amounts of text
- microcopy options
- organising research notes
- ideas for user flows or
- hypotheses to investigate later
But UX ultimately deals with real people.
Nielsen Norman Group also tested synthetic users – AI-generated people intended to simulate research participants. Their conclusion was clear: synthetic users can help with preparation and hypothesis generation, but they should not replace research with real users.
An AI model can generate a plausible answer. It cannot have the actual experience, context or behaviour of the person who will use the product.
Web design and development: speed is easier to measure
The effect of AI is already highly visible in web development.
Stack Overflow's 2025 Developer Survey found that 84% of respondents use or plan to use AI tools in their development process. Among professional developers, 51% use AI tools daily.
AI can help with:
- repetitive code
- components and variants
- debugging
- documentation
- testing
- accessibility checks
- responsive adaptations
- work between design and
- rapid
A controlled study published by Microsoft Research asked developers to complete the same JavaScript task. The group with access to GitHub Copilot completed the task 55.8% faster than the control group.
That does not mean all software development becomes 55.8% faster. The study measured one clearly defined programming task. Real products also involve architecture, integrations, security, content, testing and many decisions that cannot be reduced to a single coding exercise.
But it shows how much time a capable AI tool can save when the problem is well defined.
The problem with “almost right”
Speed has a downside.
In the same Stack Overflow survey, 46% of developers said they actively distrust the accuracy of AI output, while 33% expressed trust. The most common frustration, reported by 66%, was AI solutions that are “almost right, but not quite.”
That matters outside programming too.
An AI result can look professional and still contain:
- incorrect information
- weak typography
- inconsistent spacing
- accessibility problems
- code that works in a demo but fails in production
- a user flow based on the wrong assumptions
- a visual direction that does not fit the brand
AI makes production easier. That makes the ability to review the result more important, not less.
Why expertise still matters
It is too simple to say that AI only helps experts. Several studies show that less experienced users can also gain substantial productivity benefits.
But expertise changes what a person can do with the output.
A designer who understands typography, composition, branding and production can identify more quickly why a generated option does not work.
A UX designer who understands research methods can distinguish an interesting hypothesis from an actual user insight.
A developer who understands the system can identify code that looks correct but creates security, performance or maintenance problems later.
A study conducted with Boston Consulting Group and Harvard Business School illustrates this boundary. 758 consultants worked on tasks resembling parts of their normal knowledge work. For tasks inside what the researchers called AI's “jagged technological frontier,” participants worked more than 25% faster and completed over 12% more tasks.
But on a more complex task outside the AI system's strengths, participants using AI were 19 percentage points less likely to reach the correct answer.
The study involved consultants, not designers. The principle is still useful: AI is extremely strong at some tasks and surprisingly weak at others. The difficult part is knowing which situation you are in.
Expertise is not only about producing the result. It is also about knowing when the result deserves to be trusted.
A more realistic AI workflow
AI creates the most value when it is part of a normal professional process.
1. Understand the problem
Before opening the tool, someone still needs to understand the goal, user, brand, content, budget and constraints.
2. Use AI to explore
This is where AI can be extremely efficient: alternative directions, drafts, images, code, research questions or structural ideas.
3. Review and refine
The output must be compared with actual requirements. What works? What is generic? What is wrong? What is missing?
4. Test against reality
A design should be evaluated in its real context. UX should involve real users when the decision requires it. Code must be tested in the real system.
5. Quality-control the result
Accessibility, performance, content, visual consistency, security and brand standards still need to be checked.
This is very different from:
Prompt → generate → publish.
That workflow can be fast. It is not necessarily robust.
AI also needs clear boundaries
Professional AI use is not only about what the technology can do. It is also about what information it should be allowed to process.
Confidential client material, personal data and internal business information should not simply be pasted into open consumer tools. Organisations need approved solutions, access rules and clear routines for what can be processed where.
That is not an argument against AI. It is the same principle used with other digital tools: the right system for the right information.
A new tool in an old process
Design has always changed with tools.
Digital layout software accelerated production. Photoshop changed image work. Figma made collaboration and more accessible. None of these removed the need to understand design.
AI moves the boundary again.
The difference is that the tool can now produce the proposal itself – text, image, layout or code – rather than only helping us execute it.
That makes judgment more important.
The most interesting future is therefore not “designer versus AI.” It is designer with AI: a professional using technology to explore more, automate repetitive work and spend more time on decisions that genuinely require human understanding.
AI can produce an option in seconds.
Knowing whether it is a good option is still a different skill.