← Insights & Guides · Updated · 6 min read

Run a Churn Study From Your AI Agent: Brief to Cited Findings

By

A churn study asks former customers what happened before they left, what they tried, and why an alternative became more attractive. An AI agent can coordinate that work end to end with the people who actually used your product.

This walkthrough invites your own former customers, such as members who cancelled a meal-kit or streaming subscription in the last 90 days, so every interview is with someone who actually left.

What does the workflow look like?

Choose your own participants → create the study → send the churn brief → review the saved plan → approve and send invitations → follow interviews → retrieve the report.

What should a good churn discussion guide cover?

The guide the agent drafts in the planning step should walk each former member back through the relationship in order, because cancellation is rarely one moment. Four parts carry most of the value.

The trigger. Start with the first time the subscription stopped feeling worth it. For a meal kit, that might be a week when the box arrived and nobody cooked, a recipe that flopped, or a price change noticed on a card statement. The moderator asks what was happening in the member’s life at that point, since a move, a new job, or a new baby often explains more than any product detail.

The alternatives. Ask what the member considered instead and what they actually do now. Grocery delivery, a rival kit, a cheaper streaming bundle, or simply going back to old habits are different competitive threats, and each calls for a different response. The guide should ask how they compared options and where they heard about the one they chose.

The final straw. Separate the slow erosion from the moment they clicked cancel. Ask what happened that day, what they read on the cancellation screen, and whether anything nearly made them stay. This is where pause options, skip-a-week features, and save offers either worked or went unseen.

Win-back. Close with what would bring them back, phrased concretely: a different plan size, fewer deliveries, a lower price, a specific recipe range. The moderator ladders 5–7 levels deep on each answer, so “too expensive” turns into what the member expected to get for the money and where the gap opened.

Interviews run 20–30 minutes, which leaves room for all four parts without rushing the story.

How do you read prevalence in a churn study?

Each finding in the report states how many interviews it appeared in, written as n of N. Read that number as a measure of how common a driver is among the people you interviewed, then look at the quotes to see how it played out. A driver that appears in a minority of interviews can still be the most fixable one, for example a delivery-window problem that a scheduling change would solve outright.

Two habits keep the readout honest. First, check the denominator on every finding, since some questions only apply to members who used a feature or saw an offer. Second, ask the agent to group drivers by stage: trigger, evaluation, and final straw. A price complaint at the trigger stage reads differently from a price complaint on the cancellation screen.

Step-by-step: the actual tool calls

Step 1: Create the study

The user picks their own participants as the recruiting method before the call:

create_study({"name":"Meal-kit cancellations, last 90 days","recruiting_method":"byop"})

Step 2: Send the churn brief

customize_study({"study_id":"<study-id>","message":"Interview members who cancelled their meal-kit subscription in the last 90 days. Explore what they first signed up for, the sequence that led to cancelling, what they switched to, and what might have changed the decision."})

The agent brings the platform’s planning questions to you and leaves the cancellation reasons to the interviews themselves. Once the plan is saved, get_study shows it in full for your sign-off.

Step 3: Invite the members who cancelled

create_participants({"study_id":"<study-id>","participants":[{"email":"participant@example.com"}]})

Swap in the cancelled-member list you exported. Invitations go out by email, and each person who accepts joins an AI-moderated interview on their own schedule, so interviews run in parallel as people respond.

Step 4: Follow the interviews

list_interviews({"study_id":"<study-id>","status":"completed","page":1,"page_size":20})

Each completed interview carries a quality label, so the agent can tell you how many members were invited, how many finished, and which interviews count toward the findings.

Step 5: Retrieve the report

generate_report({"study_id":"<study-id>"})
get_study_report({"study_id":"<study-id>"})

The report lays out findings, participant responses, sample profiles, and recommendations, each tied to source references. get_interview opens the transcript and recording behind any churn driver you want to hear first-hand.

What do you get out the other end?

At each step the agent does one job: it drafts the churn brief, relays the platform’s planning questions, shows you the saved plan, sends invitations once you approve, tracks completions, and pulls the report. What comes back is a set of findings, each with prevalence (n of N) and verbatim quotes linked to the source interview and, for voice and video studies, to the moment in the recording. For example output, explore the Walmart study: a User Intuition study of 43 Walmart shoppers in which 36 of 43 will shop there again even though 17 of 43 left with an unresolved obstacle. A churn readout has the same shape.

A useful churn readout separates the initial problem, the triggering event, any recovery attempts, and the final decision. Ask the agent to attach evidence to each finding and show the exceptions. A stated price objection may reflect low perceived value or a delivery problem, and the interviews show which one it was.

Each readout states the cohort and the number of interviews, so a retention change is grounded in what those members actually said.

How do you turn churn findings into a save playbook?

A save playbook pairs each churn driver with the intervention that addresses it and the moment to deliver it. Ask the agent to draft one from the report, with a row per driver: what the member experienced, the quote that shows it, the stage it happened at, and the response your team could offer.

Drivers that surface early, such as skipped boxes piling up, point to in-product nudges well before cancellation. Drivers that surface on the cancellation screen point to the offer itself: a pause instead of a cancel, a smaller plan, or a recipe change. Drivers rooted in life events, such as a move, often call for a clean exit and a well-timed win-back message rather than a discount.

Keep the source quotes in the playbook. When the retention team debates whether a pause option is worth building, the words of the members who wanted one settle the argument faster than a summary line.

How do you follow up with earlier studies?

Ask the agent to compare this cohort with last quarter’s, or to check what members said about a save offer you tried before. It answers from your earlier studies and cites the interview behind each point.

For a quarter-to-quarter comparison, the agent pulls each study’s report and opens the interviews behind the biggest changes, so you can see whether the people who cancelled changed or the reasons did.

Waves compare cleanly when the guide stays stable. Reuse the same four-part structure each quarter and add new questions at the end, so the trigger, alternatives, final-straw, and win-back answers line up across cohorts. Then ask the agent three questions: which drivers grew, which shrank after a fix shipped, and which are new. If you launched a pause option last quarter, the next wave shows whether members saw it, tried it, and still left, and why.

Why do your own customers matter?

A panelist who has used meal kits can describe what they expect from the category. A member who cancelled yours can tell you exactly what happened in that relationship. Interviewing your own former customers keeps the research tied to the actual churn event and makes findings useful for your retention and product teams.

See the research API and MCP server for AI agents, the MCP reference, and the churn analysis solution.

Note from the User Intuition Team

User Intuition provides AI-moderated qualitative research for agencies, consulting firms, and research teams. Keep your methodology and discussion guide, bring your own sample or use our 4M participant panel, and review recordings, transcripts, and evidence-linked findings. Your researchers connect the evidence to the client decision and prepare the final recommendations.

Inspect complete sample calls and a readout, then test your own brief. Starter voice interviews cost $30 with your sample or $60 with standard panel recruitment, with no monthly fee. Specialty audiences are quoted separately; incentives you arrange for your own sample are additional. See pricing or try 3 free voice interviews with your own participants.

Frequently Asked Questions

Tell the agent the churn question, which former customers to invite, and the time window. The agent drafts the study, works through the planning step with you, and shows the saved plan. Approve the plan and the invitations, and the agent retrieves completed interviews and findings, each cited to its quotes.

Export the list of members who cancelled from your CRM or billing system and give it to the agent. Once you approve the plan and the list, User Intuition sends the invitations by email, or you can share a study link yourself.

The agent creates the study, sends your brief, relays the platform's planning questions, and shows you the saved plan. For churn you choose your own former customers as participants, approve the invitations, and the agent tracks interviews and pulls the report as they complete.

Findings with prevalence (n of N) and verbatim quotes linked to the source interview and, for voice and video studies, to the moment in the recording, plus participant profiles and recommended next steps for retention.

Yes. Bring your own participants for questions about your actual customers, such as churn. You approve the plan and the recipient list before invitations go out. The vetted panel suits category questions, such as why people cancel streaming subscriptions in general.

Ask the agent to compare this cohort with last quarter's. It answers from both studies, cites the interviews behind each difference, and opens any interview you want to hear in full.
Get Started

Put This Framework Into Practice

Sign up free and run your first 3 AI-moderated customer interviews — no sales call. Panel recruiting is billed separately.

Self-serve

Launch your first study in minutes. Results in 24 hours.

See it First

Explore a real study output — no sales call needed.

No contract · No retainers · Panel results in 24 hours