Working with the AI Consultant

Koji's AI Consultant is your collaborative partner for study design — a research-savvy conversational guide that helps you go from a rough idea to a complete, publishable research brief. The more effectively you communicate with it, the better your study will be.

How the AI Consultant Works

When you create a new study, you enter a chat-based workspace. On one side, you'll see the conversation with the AI Consultant. On the other side, you'll see the research brief — a living document that the Consultant builds and updates as you talk.

The Consultant doesn't just ask you a series of fill-in-the-blank questions. It has a genuine conversation with you, probing your goals, challenging vague assumptions, suggesting approaches you might not have considered, and progressively shaping the brief into something ready to publish.

For a deeper look at the technology behind this, see Understanding the AI Consultant.

Starting the Conversation

The first message you send sets the tone for the entire design session. Here's how to make it count.

Be Specific About What You Want to Learn

The Consultant uses your opening message to understand the scope and direction of your study. Compare these two starting points:

Vague: "I want to learn about our customers." Specific: "I want to understand why enterprise customers in the healthcare vertical cancel their subscription within the first 90 days, and what we could have done differently to retain them."

The second message gives the Consultant enough context to immediately suggest a methodology, define a target participant, and start drafting relevant interview questions — including recommending appropriate structured question types like NPS scales or satisfaction ratings. The first message would require several rounds of back-and-forth just to get to the same starting point.

If you've already written a research question, paste it in as your first message. That's exactly the kind of specificity that helps.

Share Context Early

Don't make the Consultant guess about your situation. In your first few messages, consider mentioning:

• What you already know — prior research findings, analytics data, hunches from your team
• What triggered this study — a drop in metrics, a product launch, a strategic question from leadership
• Who your participants are — their roles, company size, technical sophistication, geography
• Constraints you're working with — timeline, budget, number of participants you can realistically recruit

You can also upload context documents like previous research reports, product specs, or customer feedback summaries. These give the Consultant additional background to work with.

Conversation Strategies That Work

Ask for Alternatives

One of the most powerful things you can do is ask the Consultant to show you options. Instead of accepting the first suggestion, try:

• "Can you show me how this study would look with a Jobs to Be Done approach instead?"
• "What would the interview questions look like if we focused on the onboarding experience rather than the overall product?"
• "Give me three different ways to phrase that question."
• "Should this be a scale question or an open-ended question? Show me both."

The Consultant is great at generating variations, and comparing options often helps you discover what you really care about.

Challenge and Push Back

The Consultant isn't precious about its suggestions. If something doesn't feel right, say so:

• "That question feels too leading — can you make it more neutral?"
• "I think we're missing the emotional dimension here. Can we add questions about how users feel during this process?"
• "This target participant description is too broad. Let's narrow it to people who signed up in the last 6 months."
• "Can we add a satisfaction scale question after the open-ended section so we get a benchmarkable metric?"

Direct feedback leads to better results. The Consultant adapts quickly when you tell it exactly what to change.

Iterate in Layers

Don't try to get everything perfect in one pass. A productive session usually follows this arc:

1. Big picture first — describe your research goal and let the Consultant suggest an overall approach
2. Methodology and participant — discuss who you want to talk to and what framework fits best
3. Interview questions — refine the specific questions, their order, and the probing strategy
4. Structured question design — decide which questions should capture quantitative data via structured question types (scales, choices, rankings) and configure probing depth
5. Fine-tuning — adjust wording, add or remove questions, check for bias

Each layer builds on the previous one. If you try to jump straight to wordsmithing individual questions before agreeing on the methodology, you'll end up going in circles.

Use the Brief as a Reference Point

As you chat, keep an eye on the research brief panel. The Consultant updates it in real time based on your conversation. If you notice something in the brief that doesn't match what you discussed, call it out:

• "The problem statement in the brief doesn't quite capture what we just talked about. Can you update it?"
• "I see the brief has 12 questions — that feels like too many for a 20-minute interview. Can we trim it down?"
• "Can we change that open-ended question about satisfaction into a 1-10 scale with anchor probing?"

Thinking of the brief as your shared canvas helps keep the conversation productive.

Things to Avoid

Don't Be Passive

The Consultant is a collaborator, not a vending machine. If you just say "design a study for me" and accept whatever comes back, you'll miss the opportunity to shape it into something truly useful for your specific context. The best studies come from active back-and-forth.

Don't Overcomplicate Early

Resist the urge to specify every detail upfront. Let the conversation unfold naturally. You can always add complexity later — it's much harder to simplify an overloaded brief.

Don't Ignore the Methodology Discussion

When the Consultant suggests a methodology — Mom Test, Jobs to Be Done, Customer Discovery, Exploratory, Lead Magnet, or another approach — take a moment to understand why. Each methodology shapes the questioning style, follow-up strategy, and analysis approach differently. If you're not sure which methodology is right, ask for an explanation or read our methodology guide.

Don't Skip the Review

Before publishing your study, read through the entire brief carefully. The Consultant builds the brief incrementally, and sometimes earlier sections need a tweak to align with decisions you made later. You can always edit the brief manually or ask the Consultant to make changes.

Revisiting and Refining

You can come back to the Consultant at any time before publishing. If you've stepped away and have new thoughts, simply reopen the study and pick up the conversation. The Consultant remembers everything you've discussed.

Some researchers find it helpful to draft the study in one session, sleep on it, and come back with fresh eyes the next day. The Consultant is there whenever you're ready.

Getting the Most Out of the Collaboration

Think of the AI Consultant as a research-trained colleague who has unlimited patience, no ego, and extensive knowledge of qualitative methodologies. It works best when you:

• Bring your domain knowledge (you know your users and your business)
• Let it bring methodological expertise (it knows how to structure studies and design structured questions)
• Meet in the middle through honest, iterative conversation

When the conversation is done, you should have a research brief that you feel confident about — one that reflects your goals, speaks to your participants, and asks the right questions in the right way, with the right mix of open-ended exploration and structured data capture.

Related Articles

• Understanding the Research Brief — what the brief contains and how to review it
• Editing the Brief Manually — making direct changes to the brief
• Structured Questions in AI Interviews — designing quantitative questions within your study
• Choosing a Methodology — understanding the available research frameworks
• Publishing Your Study — going live with your study

Further reading on the blog

• How to Write User Interview Questions That Get Real Answers — Most interview questions are too narrow, too leading, or too hypothetical. Here is a practical guide to writing questions that unlock genuin
• Koji vs Strella: The Real Comparison of AI-Moderated Research Platforms (2026) — Strella hit $1.6M ARR with enterprise pricing. Koji wins on transparency, six structured question types, and accessibility. Honest 2026 comp
• 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

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