Running an Interview

The participant experience: phases, live spec panel, MoSCoW chips, and completion.

This page describes the interview experience from the participant's point of view. Participants are typically non-technical stakeholders - product managers, founders, or domain experts - who have been granted interview access to a project by a manager or org admin. See Access Control.

Opening the interview

Navigate to the project's Interview tab in the Rensei web app:

https://app.rensei.ai/[orgSlug]/[projectSlug]/interview

The page loads a two-panel layout:

Left panel - the streaming chat interface where you converse with the AI agent.
Right panel - the live spec panel, which shows the User-Spec as it is being built in real time.

If the feature has not been enabled for the project, the page shows a message directing you to an org admin. See Configuration.

The six-phase flow

The interview opens with a short welcome message explaining the process, then immediately begins Phase 1. Each phase ends when the agent determines it has gathered enough information - it summarizes what was captured in a phase summary card in the right panel, then advances automatically.

Phase 1: Context

The agent asks about the project or product as it exists today and what has changed to prompt this piece of work. Representative questions:

What does this project or product do right now?
What recent event, opportunity, or pain point is driving this work?
Are there any existing systems this needs to integrate with or replace?

You can answer in plain sentences - the agent handles structure. There is no minimum or maximum length.

Phase 2: Users

The agent shifts focus to the people who will be affected by the work. Representative questions:

Who are the primary users of this feature?
What are they trying to accomplish?
What frustrates them today about the current solution?

If you are unsure about certain users or edge cases, say so - the agent marks those areas as assumptions in the spec.

Phase 3: Core Problem

The agent asks you to articulate the problem in a single, focused statement. It will push back gently if the statement is too broad or contains solutions rather than problems. This phase often produces the most iteration - that is by design.

Phase 4: Features (MoSCoW)

The agent surfaces features one at a time based on what was discussed in earlier phases. For each feature you assign a priority using the MoSCoW chips in the chat interface:

Chip	Meaning
M - Must have	Required for launch; the work is incomplete without this.
S - Should have	High value; include if timeline allows.
C - Could have	Nice to have; defer if constrained.
W - Won't have	Explicitly out of scope for this work.

You can also type a priority in text - the agent parses it correctly. The live spec panel updates with each assignment so you can see the full MoSCoW table growing in real time.

Phase 5: Constraints

The agent asks about constraints that will bound the work:

Timeline and budget expectations
Platform or technology constraints (must work on mobile, must use existing auth system, etc.)
Compliance or regulatory requirements
Integration or dependency constraints

Unanswered questions about constraints are recorded as open items in the spec.

Phase 6: Success Criteria

The agent asks what done looks like - specifically, what measurable signals would confirm the work was successful. Representative questions:

How will you know this feature is working as intended?
What metrics or outcomes would satisfy the original goal?
Are there acceptance tests or user scenarios that must pass?

At the end of this phase the agent writes the final section of the spec and presents the completion ceremony.

The live spec panel

The right panel updates continuously as the interview progresses. Each phase summary card appears when the agent marks a phase complete:

Phase summary cards are read-only during the interview.
The MoSCoW table in Phase 4 is the only section that shows individual item chips as you assign them.
The full spec becomes downloadable as Markdown at the completion ceremony.

You can scroll the spec panel at any time during the interview without interrupting the chat.

Completion ceremony

After Phase 6 the agent presents a summary of the full User-Spec and confirms that all six phases are complete. The chat panel shows two actions:

Review spec - opens the spec in full-screen read mode. You can scroll, copy, or download it as Markdown.
Send to Rensei - submits the spec to the platform.

What "Send to Rensei" does

Clicking Send to Rensei:

Creates a new issue in the project's connected tracker (Linear or GitHub Issues) with the User-Spec as the issue body.
Sets the issue to the Icebox state in Linear (or applies the sdlc/icebox label in GitHub Issues).
If the project has the interview-v1 template or the default SDLC active and subscribed, the research agent dispatches automatically within seconds.
Redirects you to the project's issue view so you can follow progress.

The research agent runs in the background - you do not need to stay on the page. You will receive a notification in Linear or GitHub Issues when the research brief is ready for review.

During the interview

Correcting an answer

If you realize a previous answer was wrong or incomplete, say so in the chat. The agent acknowledges corrections and updates the spec. It does not require you to restart the phase.

Skipping a question

You can type "skip" or "I don't know" at any point. The agent marks the item as an open question in the spec and moves on. Open questions are surfaced to the research agent as areas needing investigation.

Abandoning an interview

Closing the browser tab or navigating away pauses the interview. Your progress is saved. Returning to the same interview URL resumes from the last completed phase. Incomplete interviews (those that never reach completion) are not submitted to the pipeline.

Overview - what interviews are and how they fit in the pipeline
Access Control - who can start an interview on a project
Configuration - enabling the feature and choosing a model profile
Interview Template - the workflow that routes the completed spec into the SDLC

On this page