Koji for Product Designers

Answer first: Most product designers don't have a researcher on call, so they ship designs informed by gut feel, internal opinions, and the loudest sales-call anecdote. Koji is the AI-native user research platform that lets designers run real, moderated user interviews in hours instead of weeks — share a link, the AI conducts the conversation, follow-ups happen automatically, and the analysis lands as a research-grade report you can pull into a Figma deck. With tools like Koji, "I'll talk to five users this week" stops being aspirational and becomes a Tuesday.

This page is the playbook for design teams: how to scope a study, what question types to use for concept and usability work, how to recruit, and how to turn 20 interviews into a defensible design decision.

The designer's research problem

Most in-house design teams sit in one of three states:

• No researcher on staff. The designer is expected to do their own research. Calendar Tetris and recruiting eat 80% of the hours. Most designs ship without research.
• One researcher across many designers. Studies queue. Designs ship on instinct or get blocked.
• Researcher available, but only for "big" work. Daily design questions — "does this label make sense," "is this onboarding flow clear" — get answered by Slack polls or hallway conversations.

Koji is built for the first two patterns and complements the third. The point isn't to replace a researcher when you have one — the point is that designers shouldn't have to wait for one when the question is small, fast, and tactical.

What Koji gives designers

• AI-moderated interviews. You scope the questions; Koji's AI consultant designs the interview guide. The AI interviewer runs each conversation, asks follow-ups when an answer is thin, and produces a clean transcript. No scheduling, no moderating, no note-taking. (How AI interviewers work)
• Voice or text. Voice interviews capture nuance — pauses, tone, "wait, actually" reversals. Text interviews scale to mobile-first audiences who can't talk out loud at their desk. Use both. (Voice vs text interviews)
• Six structured question types. open_ended, scale, single_choice, multiple_choice, ranking, and yes_no — so a concept test can mix "rate this 1–7" with "tell me why" without losing structure. (Structured questions guide)
• Auto-analyzed reports. Themes, sentiment, pull quotes, and per-question summaries — generated automatically once enough interviews complete. (Generating research reports)
• Concept testing built in. Drop in a Figma prototype link or static screens; the AI will ask the test tasks and probe the participant's reaction. (AI concept testing guide)

Where designers use Koji week to week

The most productive design teams we see use Koji at three points in the design cycle:

1. Before you open Figma — discovery interviews

Before designing anything, designers should know what problem they're solving and for whom. Six to ten discovery interviews — run as Koji studies — surface mental models, pain language, and adjacent tools. This is the moment when "the user wants X" becomes "the user says X because Y, and they've tried Z" — the kind of texture that survives the first stakeholder review. See customer discovery interviews for the methodology and Mom Test user interviews for how to avoid leading questions.

Koji's AI consultant generates the discussion guide from a one-paragraph research brief — designers no longer have to remember the rule against compound questions or "would you" hypotheticals. The AI applies it.

2. While you're iterating — concept and preference testing

Most design decisions reduce to "version A or version B." Koji handles both halves at once:

• A single_choice or ranking question gives you the quantitative preference.
• An open_ended follow-up captures why — and the AI probes the answer until it has substance, not "I just like it more."

For visual concepts and copy directions, the AI interviewer shows the option (image or text), asks the structured question, then conducts a short qualitative probe. Compare this to a SurveyMonkey or Typeform preference test, which gives you the vote without the explanation. The preference testing guide, first click testing guide, and messaging testing guide walk through each pattern.

3. After you ship — usability and friction interviews

Once a feature is live, the question shifts from "do users like the concept" to "can they actually complete the job." Koji is great for unmoderated usability research because the AI interviewer:

• Walks the participant through scripted tasks (usability testing guide).
• Probes the participant whenever they pause, hesitate, or use vague language.
• Captures friction in the participant's own words instead of an analyst's paraphrase.

Because the moderator is an AI, you can run 30 unmoderated usability sessions in the same week — something a single-designer team could never do live. The always-on user interviews with a 24/7 AI moderator doc covers the operational pattern.

A starter playbook: 5 studies a designer should run in their first 30 days on Koji

1. One generative interview study on the persona you're designing for (8–12 interviews). Output: a real persona doc grounded in quotes, not stereotypes. See user persona research guide.
2. One JTBD switch interview study with users who recently adopted (or churned from) a competing product (6–10 interviews). Output: the actual progress your design needs to enable. See JTBD switch interviews at scale.
3. One concept test comparing two design directions (15–25 respondents). Output: a defensible direction with verbatim "why."
4. One unmoderated usability study on the current production flow (10–20 sessions). Output: friction map.
5. One micro-study on the lowest-confidence label, button copy, or empty state in the current sprint (10 respondents, 4 questions, 90 seconds each). Output: a defensible micro-decision.

In most product orgs, completing those five studies would take a quarter. With Koji's AI moderator running interviews in parallel, designers complete them in 2–3 weeks of calendar time and ~5 hours of design-team effort.

How designers integrate Koji with Figma and design tools

• Embed prototypes in concept tests. Drop a Figma share link into the interview's structured-question prompt. The respondent sees the prototype and the AI asks the test tasks.
• Pull quotes into design specs. Once a report lands, copy the AI-extracted pull quotes directly into Figma annotations. Stakeholders trust quoted users more than paraphrased "we heard."
• Push findings to Notion. Use the Notion research integration to mirror Koji reports into a design-system or research-repository database.
• Ship findings to engineering. The Linear research integration turns each surfaced theme into a tagged Linear issue so designers and engineers see the same evidence trail.

Comparison: Koji vs the design-research tools designers already use

• Maze and UserTesting are excellent for unmoderated usability click-throughs, but their open-text capture is shallow — there's no AI moderator that asks "why" when an answer is vague. Koji combines task flows with conversational depth. See Koji vs Maze and Koji vs UserTesting.
• Typeform and Google Forms can pose questions but cannot probe. The most interesting answer in any concept test is the second sentence — surveys never get it. See Koji vs Typeform.
• Dovetail and Marvin are repositories that organize research you've already collected — they don't collect it. Koji collects, analyzes, and writes structured artifacts that can flow into either repository. See Koji vs Dovetail and Koji vs Marvin.

Cost: what designers actually spend

Koji uses a credit model. Text interviews cost 1 credit, voice interviews 3, and a report refresh 5. Plans start at €29/month (Insights, 29 credits/month) and €79/month (Interviews, 79 credits/month). Only interviews scoring 3+ on the quality gate count toward your credits — so spam or one-line answers don't burn budget. The plan comparison guide has the full feature matrix.

For most solo designers, the Insights plan covers a normal week of design research. Teams that run concept tests every sprint typically sit on Interviews.

Trust and craft

A recurring concern from senior designers: "if an AI runs the interview, does it leak the designer's biases into the probe?" Koji's AI interviewer is tuned to avoid leading and to follow the participant's language back into the next question — but the right discipline matters. The AI interview hallucinations & bias mitigation doc and the avoiding bias in interviews doc cover both the platform-level guarantees and the methodology. Bring those to your next design crit — it's the same conversation you'd have about a human moderator.

Related Resources

• Structured Questions Guide — the six question types and when to use each.
• AI Concept Testing Guide — concept testing with conversational follow-up.
• Preference Testing Guide — A/B preference with the "why" attached.
• Usability Testing Guide — unmoderated usability with an AI moderator.
• Koji for Product Managers — the equivalent playbook for PMs.
• Koji for UX Researchers — the playbook for trained researchers.
• How AI Interviewers Work — what the AI moderator does behind the scenes.