Skip to content

AI-powered usability testing for product development

How it works

From the first question to the next round of tests

One product, followed through a full cycle. AI helps prepare the study and suggests findings; your team reviews both and makes the decisions. That’s what human-in-the-loop means in TestFlow.

Try it with your product

Five steps, then test again

Example study · fictional data

  1. Goal

    Define what you need to learn

    Start a project for a physical or a digital product. Write down the product, the people who will use it, where it will be used and the question the study has to answer. That question stays at the top of the plan.

    “A self-service kiosk where commuters buy and top up transport tickets. We want to know if they can do it without help.”

  2. Draft and review

    Draft the study with AI, then review it

    From your description, TestFlow drafts the objectives, task groups, scenarios, moderator scripts and questionnaire. You can also start from a template or import a study you already have. Nothing reaches a participant until your team has reviewed it.

    What to check in the draft

    • Do the objectives answer your question?
    • Is the wording clear to your participants, without hinting at the answer?
    • Does each task fit the product and where it will be used?
    • Is the questionnaire short enough, and is SUS included if you need it?
    Task 3, Get a printed receipt, selected in the task list. The panel beside it shows the scenario, the moderator script and what to observe.

  3. Session

    Observe people using the product

    Add the participant and start the session. The moderator reads each scenario, times the task, records the outcome and notes what the participant says and does. Keep practice sessions separate from real ones, so they don’t count in the results.

    Live session with participant Emma Bauer on task 3 of 6. The outcome is set to Not completed and her answer is recorded.

  4. Evidence

    Review the evidence

    Results update as each session is saved. Flagged tasks point to likely problems, and AI can suggest findings from the results and notes. Your team checks each suggestion against the evidence and keeps, edits or discards it.

    Potential usability issues, each with a Record finding button. Get a printed receipt is flagged: high fail rate 43.8%, low success rate 56.3%, very variable times.
    Flagged automatically from the session results — one click to record a finding.
    Finding recorded for Get a printed receipt, severity High, status Accepted, with the decision: show Print receipt on the payment confirmation screen, then test the task again.
    The finding and the decision, with the task’s results as evidence. Accepted findings open the PDF report.

  5. Next round

    Improve the product and test again

    Record the decision for each finding, export the data or the report and change the design. Then duplicate the test for the next prototype and run it again to see whether the problem is gone.

What you need before your first session

The product or a prototype

Something people can actually use: a working prototype, a model, a pre-production sample or a clickable design.

Participants

A handful of people who match the users you have in mind, and their consent to take part.

A clear goal

The question the study has to answer, and what you will decide with the results.

The conditions

A place, the equipment and the time for each session, as close as possible to how the product will really be used.

Try it with your own product

Your first study can be ready to run today.

Try it with your product Sign in

What happens after you sign up

  1. Confirm your email address and sign in.
  2. Start a project: blank, from a template, from an imported file, or drafted by AI from a short description.
  3. Add participants, invite your moderators and run the first session.