Understanding the Research Brief

The research brief is the blueprint for your study — it's the document that tells Koji's AI interviewer exactly how to conduct your conversations with participants. Understanding what each section does (and why it matters) helps you create better studies and catch issues before you go live.

What Is the Research Brief?

When you work with the AI Consultant, the conversation produces a structured document called the research brief. You'll see it in the artifact panel to the right of the chat. This brief contains everything Koji needs to run your study: the problem you're investigating, who you want to talk to, how the conversation should be structured, and exactly what questions to ask — including structured questions that capture quantitative data alongside qualitative depth.

Once you publish your study, the brief becomes the instruction set for the AI interviewer. So what you see in the brief is very close to what participants will experience.

Sections of the Research Brief

Let's walk through each section, what it contains, and what to look for when reviewing.

Problem Context

This is the core of your study — a set of fields that capture what you're trying to learn and why it matters.

• Problem Statement — a concise description of what you're investigating, typically one to three sentences that capture your research question in context
• Decision to Inform — what business or product decision this research will help you make
• Hypothesis — what you currently believe to be true, which the study will test
• Success Criteria — how you'll know the research was successful
• Problem Cost — the impact of the problem if left unaddressed (e.g., revenue loss, user churn)
• Out of Scope — what this study explicitly will not cover

What to look for:

• Does the problem statement accurately reflect what you want to learn?
• Is the hypothesis testable through conversations with participants?
• Are you clear about what decisions this research will inform?

The problem context sets the frame for everything below it. If this section feels off, the rest of the brief will drift too. Don't hesitate to ask the AI Consultant to revise it, or edit it manually.

Methodology

This section identifies which research framework your study follows. You can learn about all of them in our methodology guide.

The methodology affects more than just the questions — it shapes how the AI interviewer follows up, what it probes for, and how it adapts to participant responses. Each methodology includes:

• Core Principles — the foundational rules the AI interviewer follows
• Question Patterns — what types of questions work well within this framework
• Probe Points — what the AI should dig into during follow-ups
• Anti-Patterns — what to avoid (e.g., leading questions in Mom Test)

For example:

• Mom Test interviews avoid leading questions and focus on past behavior rather than hypothetical preferences
• Jobs to Be Done interviews dig into the circumstances, motivations, and tradeoffs behind decisions
• Customer Discovery interviews focus on validating problem-solution fit through open exploration
• Exploratory interviews follow interesting threads with high probing depth for maximum qualitative discovery
• Lead Magnet interviews capture quotable statistics and benchmarkable data for public reports

What to look for:

• Does the methodology match the type of insight you need?
• If you're unfamiliar with the chosen methodology, ask the AI Consultant to explain why it recommended it

Target Participant

This section defines who should participate in your study. Rather than relying on demographics, Koji focuses on behavioral characteristics — what people have done, experienced, or are currently doing. This behavioral focus produces better participants and more relevant conversations.

The target participant section includes:

• Required Experience — what the participant must have experienced (e.g., "has evaluated at least two CRM tools in the past year")
• Behavior of Interest — the specific behavior your study investigates (e.g., "abandoned checkout in the last 30 days")
• Relationship to Problem — how the participant connects to the problem you're studying
• Screening Question — a question to verify participant fit at the start of the interview

What to look for:

• Is the description specific enough to recruit? "Small business owners" is broad; "founders of B2B SaaS companies with fewer than 50 employees who launched in the last two years" is actionable.
• Does it focus on behavior and experience rather than demographic categories? Koji deliberately avoids demographic-based targeting in favor of behavioral criteria that better predict relevant insights.
• Is the participant directly connected to your problem statement? If you're studying onboarding friction, you want people who have recently onboarded — not long-tenured power users.
• Are there any groups you should exclude? Sometimes defining who you don't want to talk to is just as important.

Interview Plan

This is the heart of the brief — the structured conversation guide that the AI interviewer will follow. It contains several components:

Opening Approach How the AI interviewer begins the conversation — setting the tone, building rapport, and easing the participant into the topic. This is configured as guidance rather than a script, so the AI can adapt naturally.

Key Questions These are the main questions for the interview. Each question has a specific type — open-ended, scale, single choice, multiple choice, ranking, or yes/no — and includes probing configuration that controls how deeply the AI follows up. For a full guide to question types and their configuration, see Structured Questions in AI Interviews.

Topics to Explore High-level themes the AI should cover during the interview, giving it flexibility to follow interesting threads within the defined scope.

Behaviors to Probe Specific actions or decisions the AI should dig into when participants mention them — for example, "probe the decision-making process when participants describe switching tools."

Guardrails Instructions about what the AI should avoid — topics to steer away from, types of questions not to ask, or boundaries to respect during the conversation.

Closing Approach How the AI wraps up the interview — giving participants a chance to share anything that wasn't covered and ending the conversation gracefully.

What to look for in the interview plan:

• Do the questions flow in a logical order? Would they feel natural in a real conversation?
• Are there too many questions? A 20-minute interview typically supports 5 to 8 key questions with probes. More than that, and you risk rushing through important topics.
• Is there a good mix of open-ended and structured question types? Open-ended questions build rapport and depth; structured questions anchor key metrics.
• Do the probes dig into the right things? Probes should uncover motivations, emotions, and specifics — not just ask for more detail.
• Is the language appropriate for your participants? A study about developer tooling should sound different from one about consumer health apps.

Interview Mode

The interview plan also specifies the interaction mode:

• Structured — the AI follows questions closely in order
• Exploratory — the AI follows interesting threads freely within the defined topics
• Hybrid — the AI starts structured and goes exploratory when it hits something interesting (this is the default)

The mode, along with estimated duration, shapes how the AI paces the conversation.

How to Read the Brief Critically

Reading a research brief is a skill. Here are some questions to ask yourself as you review:

The "So What" Test: For each interview question, ask yourself: "If a participant answers this, will the answer help me make a decision?" If not, the question might not be pulling its weight.

The Flow Test: Read the questions in order, out loud if possible. Do they feel like a natural conversation, or do they jump around? Abrupt topic changes can make participants feel disoriented.

The Bias Check: Look for questions that assume a particular answer. "What frustrated you about the onboarding?" assumes frustration existed. "How would you describe your onboarding experience?" is more neutral.

The Length Check: Count your key questions and estimate how long each one might take, including follow-ups. If your math says the interview will run over your target time, trim before you publish.

The Participant Check: Imagine a specific person from your target participant description reading each question. Would they understand it? Would they have something meaningful to say? If the language is too jargony or the questions are too abstract, simplify.

The Quantitative Check: Review your structured questions. Are you capturing the metrics you need for reporting? Do scale questions have appropriate ranges and meaningful endpoint labels? Will your structured data answer the quantitative questions your stakeholders care about?

Making Changes

If anything in the brief needs adjustment, you have two options:

1. Chat with the AI Consultant — describe what you want to change, and it will update the brief. This is great for significant structural changes or when you want suggestions.
2. Edit the brief directly — use the structured editor to modify any section. The editor is organized into tabs (Problem, Participant, Approach, Questions) for easy navigation. See our guide on editing the brief manually for details.

Both approaches update the same document, so you can switch between them freely.

When the Brief Is Ready

Your brief is ready to publish when:

• The problem statement clearly captures your research question
• The methodology matches the type of insight you need
• The target participant is specific and recruitable based on behavioral criteria
• The interview questions flow naturally and avoid bias
• The mix of open-ended and structured questions captures both qualitative depth and quantitative data
• The number of questions fits your target interview length
• You've read through the whole thing at least once and feel confident about it

Once you're satisfied, head over to Publishing Your Study to make it live.

Related Articles

• Editing the Brief Manually — making direct changes using the structured editor
• Structured Questions in AI Interviews — designing quantitative questions within your study
• Publishing Your Study — going live with your study
• Choosing a Methodology — understanding the available research frameworks
• Writing a Research Question — crafting the question that drives your study

Further reading on the blog

• Customer Interview Questions: 50+ Templates for Discovery, Churn, and Win/Loss (2026) — The template is not the bottleneck — conducting the interview at scale is. Here are 50+ customer interview questions organized by use case,
• How to Run AI-Powered Customer Interviews at Scale — Learn how to conduct effective customer interviews at scale using AI. This comprehensive guide covers everything from planning and question
• The Mom Test for Customer Interviews: How to Ask Questions That Get Real Answers (2026) — The Mom Test is the founder's guide to customer interviews that actually work. This complete guide covers Rob Fitzpatrick's three rules, rea

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