How do you measure whether AI is working in a design studio?

In short: Measure five things before AI enters the studio and again after two live projects: time to a client-ready concept round, number of design directions explored per project, hours of repetitive production per project, consistency of output across the team, and the share of the team using the new workflows weekly. If the first three move clearly and the last two do not fall, AI is working. If only enthusiasm moved, it is not.

Why most studios cannot answer the question

Ask a studio six months after “adopting AI” whether it worked and the answer is usually a feeling: the designers like the tools, or they have drifted back to the old way. Feelings are not a basis for changing how a practice works or deciding to invest again. The fix is inexpensive: take a handful of baselines before starting, on real projects, and compare later.

The five baselines

1. Time to a client-ready concept round

Count the working hours between receiving a brief and having a presentation the lead designer would show. Measure it on two recent projects to get a fair number.

2. Directions explored per project

How many distinct directions did the team seriously consider before choosing? This is the quality metric hiding inside a speed story.

3. Repetitive production hours

Count hours spent on work nobody would call design: first-pass renders, image variations, resizing, material boards, presentation formatting and rewriting descriptions.

4. Consistency across the team

Do three designers produce work that reads as one studio? Score a sample of output before and after against the studio's own language document. AI done badly makes this worse; AI done well makes it better.

5. Adoption

Track the share of the team that used the new workflows on real work in the past week. This is the number that predicts whether the other four will still be true in a year.

What a good result looks like

After two live projects, the direction should be visible: concept time down, directions explored up, repetitive production reduced, consistency unchanged or improved, and most of the team using the workflows weekly. Exact percentages depend on the studio, the work and the baseline; the important thing is to compare like with like.

What not to measure

Number of images generated rewards noise. Hours spent “using AI” measures a cost rather than a result. Client reaction to the word “AI” says little about the quality or efficiency of the work.

How to run the measurement without a project office

A studio does not need software for this. One spreadsheet, five rows and two columns—before and after—filled in by the lead designer at the end of each project. The discipline of writing the number down is what makes the conversation about facts.

Frequently asked questions