How to Do User Research Without a UX Researcher

Most solo founders skip user research. The justification is usually the same: no budget for a researcher, you already talk to customers, and shipping is more important than studying.

The problem is that "talking to customers" without structure produces confirmation bias, not data. You remember the feedback that matched what you already believed. You miss the pattern that appears in 7 of 10 conversations because no one catalogued it.

User research doesn't require a $90,000/year hire. The UX Researcher agent in the Design department handles the framework. You handle the conversations.

What Is User Research, Actually?

User research: The systematic process of collecting data about how real people think, decide, and behave when interacting with a product. It includes interview design, participant screening, usability test frameworks, and pattern synthesis from qualitative feedback.

The word "systematic" is the part most founders miss. Talking to three users after a rough week isn't research. It's anecdote collection. Research has a repeatable method, defined participants, and a synthesis process that separates signal from noise.

How to Run User Research as a Solo Founder

The UX Researcher agent handles everything except the actual conversations. You get structured input and output. The 30-minute calls with real users stay with you.

1. Define one research question: Start specific, not general. "Why do users stop using the product after day 7?" is a research question. "What do users think?" is not. A clear question determines which participants you need and which questions belong in the script.

2. Generate a screener survey: Paste your research question and target user description to the agent. It produces a 5-question screener to find participants who match your segment. This filters out people who don't represent the users you care about most.

3. Get a structured interview guide: The agent writes an 8-12 question script with open-ended questions, follow-up probes, and timing guidance. It avoids leading questions by default. A leading question ("Did you find onboarding confusing?") pushes people toward yes. A neutral one ("Walk me through your first 15 minutes with the product.") surfaces actual behavior.

4. Run 6-8 interviews yourself: Schedule 30-minute video calls. Read from the script. Record with permission. Don't improvise or steer toward what you hope to hear. Six interviews surfaces major patterns. Eight is enough to act on them with confidence.

5. Feed your notes back to the agent: After all calls, paste your notes in with context per interview ("Interview 3, freelance designer, 2 months on free plan"). The agent synthesizes across sessions: frequency counts per theme, contradictions between user groups, low-frequency findings that may signal edge cases.

6. Get a decision-ready synthesis: The output is grouped by insight category, not by interview. "Feature discoverability: mentioned negatively in 5 of 7 sessions, all free-plan users" is something you can act on. "User 4 said they couldn't find the export button" is not.

A Real Example

Say you're building a budgeting tool and free trial users aren't converting. You'd brief the UX Researcher agent: "Free trial users, didn't upgrade, need to understand what stopped them."

The agent produces a screener for users who fit that profile and an interview guide covering: setup experience, the moment they considered upgrading, what they compared you against, and what would have changed their decision.

After 7 calls, you paste back your notes. The synthesis shows 5 of 7 users didn't realize the tool could connect to their bank account. They thought it was manual-entry only. That's a positioning problem, not a pricing problem. You fix how you describe the product during the trial, not the price.

One research cycle, $10.25/month for the Design department. The same finding through churn analysis would have taken 3 months of data and still wouldn't have explained the reason.

Common Mistakes

Interviewing your happiest users. They confirm what's already working. You need users who left, didn't convert, or are on the fence. The screener step exists to find the right participants, not the most accessible ones.

Rewriting the interview script. The agent's questions are neutral by design. When founders improvise, they unconsciously ask leading questions. Stick to the script. The follow-up probes are already built in for when a user says something interesting.

Stopping at 3 interviews. Three calls give you three stories. At three, you can't tell if a finding is a pattern or a coincidence. Six to eight sessions is the minimum for a decision you'd build a feature on.

Waiting for a "proper" research sprint. You don't need a research plan, a Figma prototype, or three weeks blocked off. One research question, one week of calls, and one synthesis session is enough to change how you prioritize.

What the Agent Handles vs What Stays With You

The UX Researcher agent produces: screener surveys, interview scripts, synthesis documents, frequency analysis across sessions, and insight categorization by theme.

You handle: recruiting participants (ask directly in your product, via email, or through community posts), running the calls, and deciding which insights to act on.

The agent won't tell you what to build. It tells you what users actually said, how often they said it, and where the patterns are. That's the input. The decision is yours.

Bottom Line

User research that gets skipped doesn't disappear. It shows up later as features nobody uses, pricing that doesn't convert, and onboarding flows that lose half your trials. Running it properly takes about 6 hours across a week: one hour to set up, four hours in calls, one hour reviewing synthesis output.

The Design department is $10.25/month. For a solo founder making product decisions without a research team, it's the difference between building on assumption and building on data.

---

Ready to put this into practice? Browse the departments and start with whichever handles your biggest current bottleneck.