Jordan Brinkman is the lead UX researcher at ERGO NEXT Insurance, where he leads end-to-end qualitative and quantitative discovery across product and marketing. His article for The ResearchOps Review introduced the second-layer concept in research automation: the observation that integrating AI into research workflows generates a distinct set of governance and operational questions that sit beneath the surface of the apparent efficiency gains. His work examines which parts of the synthesis process can be accelerated responsibly, which must remain human-led, and how research professionals can preserve the psychological and craft dimensions of user research as they adopt (and build) new tools.
How to AI UXR is supported by Strella, an AI-powered customer research platform that partners with you to build, moderate, and synthesise interviews, allowing you to go from question to actionable insights in just a few hours.
In This Conversation
For years, research repositories have promised to make insights reusable beyond the immediate product team. In practice, many research professionals have found them difficult to keep “alive.” This challenge typically lies in the operational overhead required to classify, analyse, cross-reference, update, and maintain repositories that become vast collections of content over time.
If you’re trying to build an AI-enabled repository, you’ll find this episode especially timely. Jordan demonstrates his “living research brain,” built with Claude Code, Markdown files, AI skills, human checkpoints, and a growing set of workflows—a system that takes research materials such as transcripts, survey data, A/B test results, and secondary research, then routes them through different analytical workflows before proposing updates to a larger research wiki.
Jordan’s research brain isn’t only an example of how AI, or Claude, can process more research material more quickly. It also demonstrates what’s possible when a researcher treats AI as part of a wider system that’s purposefully designed to depend on craft, judgement, governance, and near-constant maintenance. Though AI can identify, route, summarise, cross-reference, and draft insights, humans must still decide what’s methodologically sound and what should be allowed into the organisation’s shared memory: the wiki.
Jordan also offers a practical reminder that building with AI isn’t only fast-tracked automation; it’s also systems design—the quality of which depends on how clearly you can explain your work to AI and those around you.
The How to AI UXR Map
This series builds on the insights shared in the How to AI UXR map, a five-page map that charts key trends, helps you pinpoint your AI maturity level, and offers practical, real-world applications you can adapt to your research systems.
In this episode, we discuss:
Why AI changes the repository problem by allowing research material to be analysed, queried, cross-referenced, and reorganised over time.
How his system identifies different resource types, including transcripts, reports, survey files, and A/B test results, then routes each one through an appropriate analysis workflow.
Why the system uses Markdown files as its basic structure, and how those files become a wiki-like knowledge base that can evolve as new evidence is added.
How Claude skills help define different analysis processes, and why those skills still need to be shaped by a researcher’s methodological judgement.
Why research professionals working with AI systems may need a much deeper understanding of research craft than was previously required.
Where he places human checkpoints in the workflow, particularly before analysis and cross-referencing are allowed to become new wiki content.
Why unchecked AI outputs can pollute a research system through overgeneralisation, weak synthesis, or misplaced emphasis.
How he is thinking about team access through GitHub, so that others in the organisation can query the research brain and eventually contribute new material.
Why building AI systems creates a new maintenance burden, including logs, versioning, quality checks, token costs, context management, and the need to track what has changed between working sessions.
Partway through the episode, Priya Krishnan, the cofounder and COO of Strella, shares her take on the conversation.
Things Referenced
Andrej Karpathy is an AI researcher and educator, a founding member of OpenAI, and former director of AI at Tesla.
Claude Code is Anthropic’s agentic coding tool, which can read codebases, edit files, run commands, and work across development tools.
Claude skills are reusable folders of instructions, scripts, and resources that help Claude perform specialised tasks more consistently.
CSV files are simple text files that store spreadsheet-style rows and columns, usually with values separated by commas.
GitHub is a platform for storing, versioning, reviewing, and collaborating on files, especially code.
Markdown is a lightweight markup language for adding structure and formatting to plain-text documents.
NotebookLM is Google’s AI research and note-taking tool, designed to answer questions from sources the user uploads or connects.
RAG (retrieval-augmented generation) is an AI framework that combines search or retrieval from external sources with a language model’s generated response.
Tokens are the units of text that AI systems process as input and output, such as words, word fragments, or punctuation.
Connect with the Guests
Jordan Brinkman, Lead UX Researcher, ERGO NEXT Insurance
Priya Krishnan, cofounder and COO of Strella
How to AI UXR is supported by Strella, an AI-powered customer research platform that partners with you to build, moderate, and synthesise interviews, allowing you to go from question to actionable insights in just a few hours.
