A guide to hiring AI designers
More products now have an AI feature: a chat assistant, a writing helper, a smart search or a summary at the top of a page. These features do not behave like a normal button. The same question can bring a different answer, and sometimes the answer is wrong. Designing for that takes care. This guide explains what AI designers do, how the role differs from nearby roles and how to run the work so a person stays in charge of the result.
Two meanings of the title
People use the title for two different jobs. The first is designing AI products. The designer works out how people ask for something, how the system answers, how mistakes are shown and fixed, and how data use is explained. The second is designing with AI tools. The designer uses generated drafts, summaries or image ideas as raw material in a normal design process, and checks them before anything is shared. Many people do both. Decide which one you need before you start, because the interview questions differ.
How it differs from other design roles
A UI/UX designer designs screens and journeys for any product and is the right hire when your product has no AI features. A user flow designer maps paths through a product, and this work still matters in an AI product. An AI engineer builds the features and a machine learning engineer works on the models and data. The AI designer sits between the person and the system and decides how the two talk to each other.
What good AI product design includes
- Clear limits: users can tell what the feature does well and what it does not.
- Visible sources or reasons where they help people check an answer.
- A graceful path for wrong, slow or blocked answers.
- Simple ways to correct, rate or report what the system said.
- A human review step before anything important happens.
- Short, honest notes on how data is used, with real choices for the user.
None of these needs advanced maths. They need careful thought about people, which is why design matters so much here.
Settle these before you hire
- Is the AI feature live, in testing or still an idea?
- Which kind of help do you want: product design, AI-assisted design work or both?
- Who supplies the model and who decides what it may say?
- Which tools may the designer use, and what data may never go into them?
- Who signs off on text and visuals that a tool helped to draft?
How to read an AI design portfolio
Screens of a chat window are easy to make and tell you little. Look for work that shows decisions: what happens when the answer is wrong, how the design handles a vague question and what changed after users tried it. Ask the designer how they test with real people. Be careful with portfolios that praise a tool and say nothing about review, because the review is where the skill lives.
What to ask a candidate
- How do you show that an answer might be wrong without making people distrust the whole feature?
- Walk us through the error states you designed for an assistant. What did you learn from testing?
- When you draft with an AI tool, how do you check and change the result?
- What would you want to know about our data before you design a feature that uses it?
- Describe a time you removed an AI feature or a step because users did not need it.
A strong answer is specific, admits trade-offs and shows concern for the user. Be wary of someone who says the AI will handle it.
Set a short paid task
Give each finalist a small task from your own product. You could ask for a one-page plan for a feature that suggests text, a set of screens for a wrong answer and how the user fixes it, or a short review of an existing assistant with three fixes. Pay for the work. Compare how clearly the person thinks about trust, errors and privacy, and how well they explain it.
Match the person to the stage
If the feature is still an idea, look for a designer who tests with quick mock-ups and is comfortable saying no. If it is live, look for someone who can read feedback, spot patterns in complaints and improve the weak screens. If your goal is a better design process, look for a person who already uses AI tools with care and can show how they review the results.
Accounts, access and rules for tools
Write down which AI tools your designer may use and what may never be pasted into them, such as customer data or unreleased plans. Keep tool accounts and design files in your company name. Share access to the design files, the research notes and the board, and no more than the work needs. A weekly short note and a monthly review of decisions keep everyone informed.
Designer, engineer and writer together
AI features go better when the designer talks to the engineer early. The engineer can say what the model does reliably and where it fails, and the designer can shape the screens around that. If your product has an assistant with a personality, include a writer or the designer who owns the tone. A short regular review of real conversations shows the whole team how the feature is doing.
Common mistakes
- Designing the happy path only, and finding out about errors from angry users.
- Hiding the fact that an answer was generated.
- Using tool output without checking it for facts, tone and fairness.
- Collecting more data than the feature needs.
- Letting a contractor hold the only copy of your files or tool logins.
Need the people who build the features? Look at our pages for an AI engineer and a machine learning engineer.















