Customer Discovery Interviews at Scale — How to Talk to 100 Customers in a Week

The Short Answer

Customer discovery is the most important and most neglected activity in product development. Teams know they should talk to customers regularly, but the logistics of scheduling, moderating, and analyzing interviews limits most teams to 3-5 conversations per month. AI-powered interviews remove these constraints — letting you run 50-100 customer conversations per week without a dedicated research team.

---

Why Customer Discovery Fails at Most Companies

The Math Does Not Work

Traditional customer discovery requires:

• Recruiting: 2-5 days to find and schedule participants
• Moderating: 45-60 minutes per interview (plus 15 min prep)
• Note-taking: Real-time or post-session transcription
• Analysis: 2-3 hours of synthesis per interview hour
• Reporting: Hours compiling findings for stakeholders

For a 10-interview study, that is approximately 40-60 hours of researcher time spread over 2-3 weeks. Most product teams cannot afford this on a regular cadence.

The result: 72% of product managers say they make decisions without sufficient customer evidence. Features ship based on assumptions, HiPPO (Highest Paid Person's Opinion), or competitor mimicry — not customer understanding.

The Sample Size Problem

With traditional methods, most teams interview 5-8 customers and call it done. Research suggests you need 12-15 interviews to reach thematic saturation for a single customer segment. For multi-segment research, multiply that number.

With 5 interviews, you are hearing anecdotes. With 50, you are identifying patterns. AI interviews make the larger sample economically viable.

---

How AI Enables Discovery at Scale

Step 1: Frame Your Discovery Question

Great discovery starts with a clear question — not about your product, but about the customer's problem. Use Koji's research question framework:

Weak: "Do customers like our onboarding?" Strong: "What are new users trying to accomplish in their first week, and where do they get stuck?"

Weak: "Would customers pay for Feature X?" Strong: "How do teams currently handle [problem X solves], and what have they already tried?"

The AI consultant helps refine your question into a structured research brief with methodology, themes, and probing guidelines.

Step 2: Choose the Right Methodology

For customer discovery, two methodologies excel:

The Mom Test — Best for early-stage validation

• Focuses entirely on past behavior and real experiences
• Prevents you from pitching your solution during the interview
• Surfaces whether a problem is real (people spend time/money trying to solve it) or hypothetical

Jobs-to-be-Done — Best for understanding switching behavior

• Reveals the progress customers are trying to make
• Identifies the push (current frustration) and pull (desired outcome) forces
• Uncovers the anxiety and habits that prevent switching

Select your methodology when setting up your study. The AI interviewer will adapt its questioning approach accordingly.

Step 3: Scale Your Conversations

Instead of scheduling 5 calls over 2 weeks, share your interview link broadly:

• In-app: Trigger interviews after specific user actions or milestones
• Email: Send to customer segments (trial users, power users, churned users)
• Website: Embed the interview widget on feedback or support pages
• Social: Share in customer communities, Slack groups, social channels
• CSV import: Upload a participant list for targeted outreach

Interviews happen asynchronously — participants talk to the AI whenever convenient. No scheduling. No time zone coordination. No researcher present.

Step 4: Let AI Analyze the Patterns

As interviews complete, Koji automatically:

• Generates transcripts for every conversation
• Assigns quality scores to filter low-effort responses
• Identifies themes and patterns across all interviews
• Extracts key insights with supporting quotes
• Produces a shareable research report

With 50 interviews, the AI surfaces patterns that would take a human analyst days to identify — common pain points, frequently mentioned competitors, recurring workflow gaps, and emergent needs you never thought to ask about.

---

Discovery Use Cases

Problem Validation

Question: "Is this actually a problem worth solving?"

Run 20-30 Mom Test interviews with your target segment. If fewer than 40% describe actively trying to solve the problem (spending time, money, or effort), the problem may not be significant enough to build for.

Segment Discovery

Question: "Which customer segment has this problem most acutely?"

Run interviews across 3-4 potential segments. Compare theme intensity, problem severity, and willingness to solve across segments. The segment where the problem is most painful and people are already paying for workarounds is your beachhead.

Competitive Switching

Question: "Why do customers choose us over alternatives (or vice versa)?"

Run JTBD interviews with recent switchers — both to and from your product. Map the forces of progress: what pushed them away from the old solution, what pulled them toward the new one, what anxieties almost stopped them, and what habits they had to break.

Pre-Launch Validation

Question: "Does our solution concept actually address the problem?"

Run interviews that explore the problem space first (without mentioning your solution), then present the concept in the final third. This reveals whether your solution addresses a real problem — not whether people are polite enough to say it sounds nice.

Post-Launch Learning

Question: "How are real users experiencing our product in their actual workflow?"

Target users 2-4 weeks after they start using your product. Understand what they expected, what surprised them, what they use daily, and what they have not touched. This feeds directly into iteration priorities.

---

Building a Continuous Discovery Practice

The real power of AI interviews is not one-off studies — it is continuous discovery. Teresa Torres' framework recommends weekly customer contact. Here is how to implement it:

Weekly Discovery Rhythm

For a fully automated pipeline, see Continuous Discovery with Koji MCP.

Integrating Discovery into Product Decisions

Discovery is only valuable if it influences decisions. Share research reports directly in:

• Sprint planning meetings
• Roadmap reviews
• Design critiques
• Executive updates

The insights dashboard makes findings accessible to your entire team — not locked in a researcher's notebook.

---

Measuring Discovery Effectiveness

---

Getting Started

1. Pick one question your team is debating or assuming the answer to
2. Create a study with that question as the research goal
3. Choose Mom Test for problem validation or JTBD for switching behavior
4. Send to 20+ participants via link or embed
5. Review insights and bring findings to your next team meeting

---

Next Steps

• Quick Start Guide — First AI interview in 10 minutes
• The Mom Test — How to talk to customers without being misled
• Jobs-to-be-Done Guide — Understand switching behavior
• How Many Interviews Are Enough? — Sample size guide
• Continuous Discovery with MCP — Always-on research pipeline
• MCP Workflow for Product Managers — Automate research with Claude