Do You Still Need a Designer If AI Builds Your MVP?

For a simple app, you can build the first version without a designer. What AI does not do is judge the result: what the user should see first, what to remove, and whether the screens work for someone who has never seen your product. That judgement comes from having seen many products, and it needs a person to review the output. Hire a designer when your prototype works but you cannot say why users would choose it, or when real users start to leave.

Prefer to listen? The 11-minute episode is on Apple Podcasts and Spotify. See all episodes.

Tools like Lovable, Cursor, Bolt and Replit can produce screens, a database and a login in days. That is real progress, and we use AI every day ourselves. The question is no longer "can I build it?" but "will anyone use it, and why?"

What are AI tool makers saying this week?

They say building is getting cheaper. On November 12, Bolt published a guide to MVP development. It argues that the old assumption that a credible MVP needs months of agency work and a five-figure budget "no longer holds for most standard builds". The same post says agencies still make sense for regulated data, deep integrations and products that last beyond the MVP, and that an interface which ignores a company's design system can hurt its credibility. Bolt sells the alternative, so read its cost and timeline claims as a vendor's view, not as data.

Other teams are moving in the same direction. Cursor made its Projects workspace multiplayer on November 10, so engineers, designers and PMs can share plans, docs and designs with their agents. Replit's agent can now buy supplies and run ads for a business, with the owner approving before anything is charged. The tools are doing more. The decisions are still yours.

What can AI do well, and what can it not decide?

AI is good at turning a clear instruction into working screens quickly. It is weaker where the instruction is the hard part. In our experience, the gaps fall in the same places:

That last point is the one founders miss. AI gives you the most common answer. The most common answer is fine for a login screen and costly for your onboarding, where you need to stand out. We wrote about why AI-built apps still feel amateur and how to fix them in order.

There is a recent example of why format and review matter. In a post published on November 10, RevenueCat described rebuilding its AI editor for paywalls. The first version completed what users asked for, but the designs lacked taste, the author says. The team changed the format the agent works in, from JSON patches to code, and reports 83% more positive feedback and 40% less negative feedback across about 46,000 requests (company-reported). The lesson is not "use code". It is that the same models gave very different quality depending on how the work was set up and checked.

Why does a designer's experience matter if AI can generate screens?

Because the skill that is hard to automate is looking at a screen and knowing what is wrong with it. A designer who has worked on many products has seen which patterns users understand and which they miss. AI does not have that background; it has an average of what exists. This is our view, not a measured result.

It is also why the work does not end when the screens appear. As with code, a person has to review the result. At Grona we make changes to both the UX and the UI of what comes back, very often. Reviewing is where a designer earns the fee.

What does that look like on a real product?

On Tenzio, a strength training app, the backend and training logic were already built and working. The interface was the weak part: dark, cyan-heavy cards, a flat hierarchy, and a Home screen that felt like a workout launcher rather than a daily status screen. We did not change what the product did. We changed what the user sees and in what order. That is a design decision, not a build task, and it is the kind of decision AI tools do not make for you. (The case page does not state results, so we do not either.)

When is a designer worth it, and when can you wait?

A starting point, not a rule:

1. You can wait if the app is a simple tool for yourself or a few people you know, and the cost of it feeling rough is low.

2. Hire before building if you are not sure which job the app does for the user. A decision is cheaper to change in a design than in built software; see MVP design vs MVP development.

3. Hire at the prototype stage if it works but feels generic, or if you want to show it to investors or real users. Our guide to using an AI prototype as a design brief shows how to hand it over.

4. Hire when users leave. If people sign up and do not come back, the problem is usually the flow, not the code.

If you are unsure whether to hire a partner or a freelancer, see design partner vs freelancer vs in-house designer.

What if I think AI will do all of it?

Test it. Build the product with AI, and let a designer work with AI on the same brief. Then compare which version users understand and use more. That is the honest answer to "AI will do everything": measure it on your product, with your users. If the AI-only version wins, you saved money. If it does not, you found out before launch.

What to do this week

1. Open your prototype as if you had never seen it. Write down the first thing a new user sees and the first thing they are asked to do.

2. List every feature. Mark the three you would keep if you had to cut the rest.

3. Ask someone who has not seen the app to try one task while you stay silent. Note where they hesitate.

4. Check that buttons, colours and spacing repeat the same way on every screen. Mismatches are a common sign of AI defaults.

5. Decide which question you cannot answer yourself. That is the one to take to a designer.

FAQ

Can AI replace a UX designer?

For a simple app, AI can produce screens without one. It does not judge what to show first, what to remove, or how a new user will react. A person has to review the result. We think this review is the designer's main job once AI builds the first draft.

How much does it cost to add a designer to an AI-built MVP?

It depends on scope, and we do not quote a single figure here. Our cost guide explains what drives the price.

Should I design first or build with AI first?

If you are not sure what the app should do, design first or test the idea cheaply, because changing a decision is cheaper before it is built. If the idea is clear and simple, a quick AI prototype is a reasonable way to learn.

If you have a prototype and are not sure what to fix first, our product design for startups service turns it into a design your developers can build from. Or book a free 30-minute Product Diagnosis and we will tell you where we would start.