Understanding the AI Consultant

The AI Consultant is your research design partner inside Koji. It takes your research question and helps you build a rigorous, well-structured interview plan — much like working with an experienced qualitative researcher, except it's available instantly and never gets tired of revisions.

If you've ever felt unsure about how to structure a research study, which questions to ask, or what methodology to use, the AI Consultant is built for you.

---

What the AI Consultant Does

At its core, the AI Consultant handles the part of qualitative research that usually requires the most expertise: research design. Specifically, it:

1. Reads and interprets your research goal — understands what you're trying to learn, even from a casually written description.
2. Asks clarifying questions — identifies gaps in the brief and asks targeted follow-ups to fill them.
3. Recommends a methodology — suggests a qualitative framework that fits your research type.
4. Drafts a research brief — creates a structured document defining objectives, audience, themes, and success criteria.
5. Builds an interview plan — generates a complete conversation flow with opening questions, core exploration, follow-up probes, and closing.
6. Designs structured questions — creates interactive question widgets (scales, choices, rankings, yes/no) with configurable probing depth for quantitative data capture alongside qualitative conversation.
7. Incorporates context — reads any documents you upload (PDFs, text files, DOCX, JSON, MD) and weaves that knowledge into the research design.
8. Iterates on feedback — revises any part of the plan based on your input, as many times as you need.

The result is a research study that follows established qualitative methods, with questions designed to surface genuine insights rather than surface-level opinions — and structured data capture that gives you quantifiable metrics alongside conversational depth.

---

How the Conversation Works

When you create a new study, the AI Consultant opens a chat-style conversation. Here's the typical flow:

Phase 1: Understanding Your Goal

You start by describing what you want to learn. The Consultant reads your input and responds with a combination of:

• Acknowledgment — confirming it understands your goal.
• Clarifying questions — asking for details that will make the research more focused.
• Initial suggestions — early thoughts on approach and methodology.

For example, if you write "I want to understand why users churn," the Consultant might ask:

• "Which user segment are you most concerned about — new users, long-time customers, or a specific plan tier?"
• "Do you have any hypotheses about why they're leaving?"
• "What time frame are we looking at — users who churned recently, or over the past year?"

These questions aren't random. They're designed to narrow the research scope so the interviews produce actionable insights, not generic feedback.

Phase 2: Methodology Selection

Once the Consultant has enough context, it recommends a qualitative methodology. Koji's AI is trained on several established frameworks:

• The Mom Test — focuses on asking about real past behavior rather than hypothetical futures. Perfect for product validation because it avoids the trap of participants telling you what you want to hear. The Consultant structures questions around concrete experiences and specific examples.

• Jobs to Be Done (JTBD) — explores the "job" a user is "hiring" your product to do. The Consultant designs questions that uncover the situation, motivation, and desired outcome that drive adoption and switching decisions.

• Customer Discovery — a structured approach to understanding problems, needs, and existing solutions in a market. Great for early-stage research when you're still defining the problem space.

• Exploratory — open-ended conversation designed to deeply understand a lived experience. Used when you're researching a topic you know little about and need to build foundational understanding.

• Lead Magnet — designed to gather quotable statistics and data points for public reports, blog posts, or marketing content. Heavy on structured questions (scales and choices) for chartable data, with open-ended questions for pull quotes.

You don't need to be familiar with any of these. The Consultant explains why it's recommending a particular approach and what it means for the interview structure. If you have a preference — maybe you've used JTBD before and want to stick with it — just say so.

For a deeper comparison, see Choosing a Methodology.

Phase 3: Research Brief

The Consultant drafts a research brief that captures everything agreed upon:

• Research objectives — clearly stated goals for the study.
• Target audience — who should be interviewed and why.
• Key themes — the main areas the interviews will explore.
• Methodology — the chosen framework and how it applies.
• Success criteria — what good interview data looks like for this study.

This brief is important because it becomes the guiding document for the entire study. The interview questions, the AI interviewer's behavior, and the analysis framework all flow from this brief.

Review it carefully. If something feels off, tell the Consultant. It's much easier to adjust the brief now than to realize after 10 interviews that the questions were pointed in the wrong direction.

For more on the brief structure, see Understanding the Research Brief.

Phase 4: Interview Plan and Structured Questions

With the brief finalized, the Consultant generates the interview plan — the actual conversation structure that Koji's AI will follow during interviews.

The plan includes:

• Warm-up section — easy, conversational questions to build rapport and set the participant at ease.
• Core questions — the main research questions, sequenced to build on each other logically.
• Structured questions — interactive widgets that capture quantifiable data mid-conversation (see below).
• Probing prompts — follow-up directions the AI should explore based on participant responses.
• Transition logic — how the conversation moves between topics naturally.
• Closing section — questions that capture final thoughts, reflections, and anything the participant wants to add.

Structured Questions: Conversational Depth Meets Quantitative Rigor

Koji combines the depth of conversational AI interviews with interactive structured questions — giving you both qualitative richness and quantitative data in a single session. During an interview, participants interact with visual widgets embedded in the conversation:

• Scale questions — numeric ratings like NPS (0-10), satisfaction (1-5), or Likert scales (1-7). You configure the range and endpoint labels.
• Single choice — select one option from a list. Ideal for segmentation and clear preferences.
• Multiple choice — select all that apply. Great for feature usage, pain points, or tools.
• Ranking — drag items to order by preference or priority.
• Yes/No — binary questions for screening or confirmation.

What makes this different from a traditional survey is the probing depth. Each structured question can be configured with 0 to 3 follow-up probes. After a participant gives a rating or selects an option, the AI automatically asks follow-up questions to understand the reasoning. For scale questions, anchor probing asks what would change the score — turning a simple number into a rich insight.

This means you get clean, aggregatable data (averages, distributions, charts) alongside the conversational context that explains the "why" behind every data point. Learn more in the Structured Questions Guide.

The interview plan is adaptive. During an actual interview, Koji's AI uses this plan as a guide but responds dynamically to what the participant says. If a participant shares something unexpected and insightful, the AI will follow that thread before returning to the planned questions.

---

Giving Effective Feedback

The quality of your study design depends on how well you collaborate with the Consultant. Here are tips for making the most of the conversation:

Be Specific About Changes

Instead of "make it better," try:

• "Add a question about how they evaluated alternatives before choosing us."
• "Include an NPS scale question with a 0-10 range and anchor probing enabled."
• "The opening questions are too formal — make them feel more like a casual conversation."
• "I want to spend more time on the post-purchase experience and less on the discovery phase."

Challenge the Recommendations

The Consultant is knowledgeable, but you know your business better. If something doesn't feel right, push back:

• "I don't think JTBD is right for this — we already know the job. I want to understand satisfaction, not motivation."
• "Our users wouldn't respond well to that question. Can you rephrase it to be less direct?"

Share Context Generously

The more the Consultant knows, the better the research design. Share:

• Previous research findings that are relevant.
• Internal hypotheses your team has about the topic.
• Constraints (e.g., "Our participants will only have 10 minutes").
• Business context (e.g., "We're deciding whether to sunset this feature").

You can type this information directly in the chat or upload documents (PDF, TXT, DOCX, DOC, JSON, MD) for the Consultant to reference.

Iterate Without Hesitation

There's no penalty for multiple iterations. The Consultant doesn't get frustrated, and the quality of the output genuinely improves with each round of feedback. Most users find that 2-3 rounds of revision produce an excellent interview plan.

---

What the AI Consultant Is Not

To set expectations clearly:

• It's not a standalone survey tool. While Koji supports structured question types like scales and multiple choice, the platform is designed around conversational depth. Structured questions are embedded within AI-led conversations, not delivered as static forms. The result is richer data than either approach alone.
• It's not the interviewer. The Consultant designs the research. A separate AI system conducts the actual interviews with participants. They're optimized for different things.
• It's not a replacement for human judgment. The Consultant makes recommendations based on methodology expertise, but you should always review and adjust based on your specific context and knowledge of your audience.

---

Getting the Most from the Consultant

Experienced Koji users develop a rhythm with the Consultant that consistently produces strong research designs. Here are the patterns that work best:

1. Start with the decision, not the topic. Instead of "I want to research onboarding," try "We need to decide whether to simplify or add more steps to onboarding. I want to understand what users actually experience." This gives the Consultant a decision anchor.

2. Share what you think you know. Telling the Consultant your existing hypotheses helps it design questions that test those assumptions rigorously rather than just confirming them.

3. Mix question types intentionally. Use open-ended questions for depth and exploration, scale questions for benchmarkable metrics (NPS, satisfaction), choice questions for segmentation, and ranking for prioritization. The Consultant can help you find the right mix.

4. Read the interview plan out loud. Before publishing, read through the questions as if you were a participant. Do they flow naturally? Do any feel awkward or leading? Share that feedback.

5. Upload context. Even a simple one-page document with product context helps the Consultant generate significantly more relevant questions.

---

Next Steps

• Creating Your First Study — put this knowledge into practice by creating a study.
• Structured Questions Guide — learn how to design effective structured questions for quantitative data capture.
• Understanding the Research Brief — deep dive into the brief structure and how it guides your research.
• Choosing a Methodology — compare qualitative frameworks to pick the right one for your next study.
• Voice Interview Experience — understand how voice interviews work for participants.
• Text Interview Experience — understand how text interviews work for participants.

Further reading on the blog

• Surveys vs Interviews: When to Use Each (And When to Use Both) — Surveys give you scale. Interviews give you depth. But choosing the wrong method wastes time and produces data you cannot act on. Here is a
• Best AI Market Research Tools in 2026: The Complete Buyer's Guide — AI has fundamentally changed market research. This guide compares the leading AI market research platforms—from AI-native interview tools li
• AI-Moderated vs Human-Moderated Interviews: Which Should You Choose? — AI-moderated and human-moderated interviews each have a time and a place. Here is the honest comparison to help you choose the right approac

<!-- further-reading:blog -->