---
title: "A research copilot, from brief to final deck"
description: "Built a modular AI workflow for UX and market research, with a shared study lifecycle and different branches for different methods."
date: 2026-10-27
lang: en
translationKey: "research-copilot"
translationStatus: machine
kind: project
name: "Research copilot"
period: "Research practice"
role: "Research workflow design and modular AI skills"
status: "Research-team tool"
tags: ["research", "AI"]
related: []
sourceUrls: ["https://ux.shipulin.md/cv#now"]
canonical: "https://shipulin.md/projects/research-copilot/"
---

A study rarely ends when its most interesting part is over. After the interviews, there is still material to organise, an interpretation to develop and a conclusion to explain to the team. Before the interviews, there is a different question: which decision is the research meant to inform?

I re-platformed UX and market research onto a modular AI copilot covering that whole journey: brief, study design, transcripts, synthesis, report and final deck. The skill set became standard tooling across the team.

## One lifecycle, different methods

The central design choice is conditional branching by methodology. Interviews, surveys and usability studies share an organisational context, but place different demands on evidence and interpretation. The workflow therefore consists of modules, rather than a single instruction to “do the research”.

This makes automation a set of specific problems. One stage needs a clear question and an intended decision. Another needs organised material. A later stage connects observations to an argument for the report. Reaching the next stage does not, by itself, establish that the evidence supports the conclusion.

## My contribution

My work was to structure the research workflow and the skills around it, accounting for the method being used. It connects two parts of my practice: conducting research and building tools for researchers.

[Grounded](/projects/grounded/), which I built for open-response coding, is a separate example of working with research material. The full copilot addresses a different organisational problem: connecting the stages of a study. This is a separate level of workflow design, alongside tools for a specific research operation.

## Where the researcher remains

A well-written report can conceal a weak transition from evidence to interpretation. For me, automation is useful when it makes that transition easier to revisit: where did an assertion come from, and which question does it answer?

Covering the full lifecycle does not mean every method becomes the same text-processing exercise. A modular structure leaves room for those differences. The researcher remains responsible for framing the problem, choosing the method and interpreting the material. The modular structure gives those decisions a place within the overall workflow.

## Sources

- [Public account of the research workflow](https://ux.shipulin.md/cv#now)
- [Grounded: a separate tool for open-ended answers](/projects/grounded/)