AI can build a website. What should you still pay a designer for?

AI has changed the economics of making a first draft. For a London business commissioning a website, the useful question is what the design team contributes after that first draft exists.

A polished homepage used to be a relatively expensive thing to produce. Now you can arrive at a project meeting with several plausible versions already on your laptop. That is useful progress. It also makes an old agency pitch — pay us because we know how to assemble pages — harder to defend.

The answer is not to pretend that AI cannot design. It can help generate layouts, images, copy and working interfaces. Nor is every human-made website thoughtful or distinctive. The difference worth paying for is whether someone can turn your business context into a clear direction, explain the decisions, and take responsibility for how the finished experience works.

That distinction is appearing beyond agency sales conversations. In October 2026, Creative Bloq examined why AI companies still commission human designers for their branding. The interesting question for a business owner is not whether those companies are being inconsistent. It is what they are buying that access to powerful generation tools does not automatically provide. [1]

The brief has become more valuable than the first layout

A tool can respond to a brief. It can also help challenge one. But someone still has to establish which parts reflect the real business, which are assumptions, and which promise something the team cannot deliver.

Consider a hypothetical London consultancy moving from general business support into specialist work for growing professional-services firms. Its old site describes everything it has ever offered. A prompt asking for a premium consultancy website may reproduce that broad positioning in a better-looking package. Production has improved; the strategic ambiguity remains.

A useful design engagement starts earlier. Who should recognise themselves in the first screen? What problem makes them seek help? Which existing projects support the new positioning? Which services should become less prominent? These decisions influence page structure, imagery and the way a visitor compares options. They are not finishing touches added once the layout is approved.

For a London company, local context should enter this discussion where it changes the relationship. Perhaps clients want in-person workshops, perhaps the team serves a particular sector, or perhaps most projects are delivered remotely across the UK. A skyline photograph is not a substitute for explaining those realities. Neither is putting London into every headline.

A real example: an AI company buying creative direction

HOW+HOW's published Ideogram project includes brand strategy, visual identity, verbal identity, asset generators and website design. The studio describes imagery generated with AI and subsequently refined in Photoshop. It is a useful example of human direction and automated production working together, rather than a campaign built on rejecting the technology. [2]

The lesson for a smaller business is about the shape of the commission. Buying a coherent identity and a way to apply it is different from buying a collection of attractive outputs. One gives future decisions a reference point. The other can leave the next person choosing a new style every time they need a page or image.

This case does not prove that hiring an agency increases revenue, or that every business needs a rebrand. It shows a specific division of work: generative tools contribute to production inside a direction developed by a creative team. That is a more useful model than arguing over whether a single image was made by a person or a machine.

Ask for decisions you can inspect

“Strategy” is too easy to sell as a vague extra. If it is part of your website budget, it should produce something you can review: a focused audience definition, a hierarchy of messages, a page structure and a reason for the selected visual direction.

You should be able to trace those decisions into the design. If the priority is helping a buyer compare three services, where does that comparison happen? If reassurance is essential, which evidence appears beside the claim? If the business needs to explain a complicated process, how does the layout make that process easier to follow?

This is also a fair way to challenge a designer. “It feels more premium” is a starting impression, not a complete explanation. Ask what creates that impression, whether it suits the audience and what it costs in readability or maintenance. An unusually large heading may establish confidence. It may also push the useful explanation out of view on a phone.

The table is a commissioning framework, not a boundary between tasks only humans or only machines can perform. AI can assist throughout. The point is to identify who checks the work and what you receive, rather than paying a premium for an undefined promise of creativity.

Art direction is a system, not a mood board

A mood board can establish a conversation about atmosphere. It cannot, on its own, tell your team how to choose the next project image or build a service page six months later.

Art direction becomes useful when it sets a few repeatable rules. These might cover how photographs are cropped, how technical information is illustrated, which typographic contrasts matter and where movement is appropriate. The rules need enough flexibility for real content, but enough discipline that new material still feels like the same business.

Imagine the consultancy again. One direction could use restrained typography and annotated project documents to make its thinking visible. Another could use expressive illustration to explain complex organisational problems. Neither is inherently better. The choice should follow the offer, the available evidence and the response the business wants to create.

Commissioning this system also reduces dependence on individual taste. Instead of asking whether everyone likes a new image, you can ask whether it follows the agreed approach and helps explain the subject. That makes feedback more useful and helps avoid endless rounds of unrelated visual changes.

Faster production should change how a project is scoped

If automation reduces repetitive work, the sensible response is to reconsider the project, not protect every old production step. Some businesses can launch a smaller site sooner. Others can use the available time to improve content, explore an additional design direction or test the experience more carefully.

Ask the agency where it expects to use AI, which outputs need review and what the saving enables. You do not need a record of every prompt. You do need clarity about the final deliverables, the review process and who is responsible for errors. A proposal should explain the work in terms you can compare with another proposal.

For a straightforward site with a clear offer and good existing assets, a template or AI-assisted build may be entirely appropriate. The value of a designer could be a focused review, a stronger page hierarchy and a small set of tailored components. It does not have to mean a large bespoke commission.

For a business changing its positioning, entering another market or combining several services, more discovery may be justified. Complexity should come from the problem being solved. It should not be manufactured through extra workshops or elaborate presentation decks that never influence the website.

Development still matters at the point of use

Moving attention towards design does not make engineering irrelevant. A good visual direction can be undermined by a fragile implementation. The site must accept real content, behave sensibly across screen sizes and remain manageable after the people who built it have moved on.

A practical handover test is to ask the future editor to make an ordinary change: replace an image, add a project, update a service description. If the design breaks as soon as a title is longer than the sample, the system is unfinished. If a small edit requires rebuilding several pages, the maintenance burden belongs in the buying decision.

The same principle applies to interactions. A mobile menu, a form error and a successful submission all need an intentional design. W3C's forms guidance provides a useful reference for labels, instructions and feedback. These are parts of the experience, not technical details to discover after sign-off. [3]

Ask to review the finished journey rather than just the most impressive screen. Someone should demonstrate what happens when information is missing, when a visitor changes their mind and when the team receives the enquiry. The successful path is only part of the product you are commissioning.

A better way to compare website proposals in London

Comparing page counts and headline prices alone can hide major differences. One proposal may assume you will supply approved copy, images and structure. Another may include researching and producing them. Neither is automatically better, but they are not equivalent purchases.

Ask each team to show how it would approach one difficult decision in your project. It might be separating two overlapping services or explaining a specialist offer without jargon. You are looking for the quality of its reasoning, not a free finished design. A relevant portfolio example should help explain how the team thinks.

Then agree the review points. Approve the message and structure before judging detailed visual polish. Review a representative page with realistic content before rolling the style across everything. Check the implemented experience before launch. This gives both sides a way to identify problems while they are still relatively inexpensive to address.

Ownership matters too. Establish who controls the platform account, where the source assets live, what the handover includes and which work is covered after launch. Clear responsibilities are part of the value of a professional engagement, regardless of how much automation happens behind the scenes.

Buy a stronger decision, not a defence of old methods

The strongest argument for professional design is not that software will never catch up. That is a prediction no agency needs to make. The argument is that your business deserves a considered direction, credible content and a finished experience someone has checked against its intended purpose.

If your current site already has those qualities, you may need only targeted improvements. If the business has changed and the site no longer expresses it, generating more layouts will not settle the underlying decisions. Start with those decisions and choose the production method afterwards.

Fit Design works from London and Gdańsk across website strategy, design and implementation. If you are reviewing your next project, explore our website services and bring the current site, the change in your business and the questions your customers keep asking. Those are better starting materials than a folder of effects to copy.

Sources and editorial scope

This is Fit Design's analysis, researched on 11 November 2026. The consultancy is a hypothetical example. No conversion uplift, market survey or client result is claimed. The cover is an AI-generated conceptual illustration, not a client project.

1. Creative Bloq — Why do AI platforms want humans to design their branding?, 25 October 2026.
2. HOW+HOW — Ideogram project, the studio's account of its own work.
3. W3C Web Accessibility Initiative — Forms Tutorial.