Does AI Make Design Cheaper? What It Actually Changes

It is a fair question, and clients ask it more every month: if AI can generate screens, copy and images in seconds, why does design still cost what it costs? The honest answer is that AI has changed the work, but it has not changed most of the price, and the reason is worth understanding before you brief anyone.

What AI genuinely speeds up

This part is real, and any designer telling you otherwise is behind.

• Ideation. Dozens of directions and starting points in minutes, so exploration is no longer limited by how fast someone can draw.
• First drafts. Layout options, placeholder content and structure in minutes instead of hours.
• Copywriting. Headlines, microcopy, onboarding text and content structure, as a first pass.
• Assets. Icons, illustrations and imagery that used to be commissioned or bought.
• Research synthesis. Summarising interviews, clustering feedback and pulling themes out of survey responses.
• Donkeywork. Resizing, variants, localised versions, renaming and documentation — the unbillable grind.
• Development. Boilerplate front-end code, component scaffolding and working prototypes, which shortens the build.
• Presentation. Turning work into client-ready decks and summaries.

All of it still lands in front of a human. AI output lacks finesse and originality, and it misunderstands briefs regularly — confidently, and in ways that look correct at a glance. Someone has to give it direction, judge what came back, refine it, add the creative idea it couldn’t have, and hold the quality bar. The speed is real. The autonomy isn’t.

For simple work, the savings can be substantial. A basic marketing page or an internal tool used by ten people is genuinely cheaper to produce than it was three years ago.

What AI does not speed up

The expensive parts of a real product project are not production. They are decisions.

• Deciding what to build. AI can generate twenty ideas. Knowing which one fits your business, budget and users is the job.
• Understanding your context. Your customers, your market, your technical constraints and the politics of your team are not in any training data.
• Edge cases. Empty states, errors, offline behaviour, long text, bad data, first-time users. This is most of the real work, and it is exactly what AI output skips.
• Alignment. Getting a founder, a developer and a marketing lead to agree on one direction takes conversations, not prompts.
• Testing with real people. Simulated users are not users. Watching someone struggle is still the only reliable signal.
• Anything genuinely new. Models are strong on patterns they have seen a thousand times. Complex products, or ones with innovative or uncommon features, sit outside that training data — and on those projects a human still does most of the work, with AI reduced to a helper.
• Accountability. When a flow fails in production, someone has to know why and fix it. That responsibility has a price.

Where AI quietly costs more

Three traps show up repeatedly on projects that lean on AI too early:

1. Plausible-looking output. AI screens look finished, which makes everyone assume the thinking is finished too. It usually is not, and the gaps surface during development.
2. More to review, not less. Generating five variations is easy. Judging them, discarding four and fixing the fifth still takes a senior eye.
3. Generic results. Models produce the average of what they have seen. If your product needs to feel distinct, the average is the opposite of what you want.

So how much does it actually save?

On a typical product design project, production is a minority of the hours. Speeding it up meaningfully still only moves the total by a modest amount. My rough experience, and it varies by project:

Type of work	Realistic saving with AI
Simple marketing page or template site	30–50%
Icons, illustrations, placeholder content	50%+
Research synthesis and documentation	30–40%
Full product design with research and testing	10–20%
Design systems and complex flows	10–15%
Strategy, alignment and accountability	Close to zero

That is why a serious product project has not halved in price. The part AI accelerates was never the expensive part.

What this means for your quote

It should mean two things. First, you should not be billed for work that is now trivial: no one should charge a day for generating a set of icons. Second, you should expect the same or more attention on the parts that decide whether the product works, because that is where the value moved.

If a quote has dropped dramatically because “we use AI”, ask what was removed. Usually it is research, testing or edge cases, and you will pay for those later in development and lost customers. See why good design costs what it does for the breakdown.

How I use AI on projects

• To get inspiration and draft additional directions quickly, so more options get explored before committing to one
• Drafting copy and content that I then rewrite properly
• Summarising research and organising notes
• Producing assets and repetitive variants
• Speeding up front-end prototypes so ideas can be tested sooner

What I do not do is let it decide what gets built, skip testing with real people, or ship anything I have not reviewed line by line. Clients get the speed, not the shortcuts.

Frequently asked questions

---

Note: Sign up to my mailing list to receive useful articles like this one every once in a while.