Welcome to Research Rundown.

We have things to do.

If you’re a researcher and you’ve spent any time on LinkedIn lately, you’ve seen it: post after post about how AI is going to change market research. AI in UX research. AI in qualitative analysis. AI in pretty much everything. The takes are everywhere. The actual workflows? Suspiciously absent.

That’s the gap this newsletter is trying to fill.

I’m Shauna. I’ve spent the last decade-plus running quant and qual research across B2B SaaS, agency, and consulting work. I’m still doing it. Which means I’ve also spent the last couple of years actually using AI in market research — not theorizing about it on a stage, but running studies, building workflows, debugging the parts where AI breaks, and figuring out what’s worth the hype and what’s just vendors selling vibes.

This is what Research Rundown is going to be about: AI for researchers, written by someone still doing the research.

Why this exists

Most “AI in market research” content right now is written by people who haven’t run a study in five years. You can tell because the advice is either too general to be useful (“AI will transform insights!”) or too vendor-shaped to be trusted (“here are five tools you should pay for”). Neither of those help you on a Monday morning when you’ve got a deadline and a survey that won’t behave.

I want this to be the place that does help. Real workflows. Real prompts. Real takes on what’s working in AI for UX research, quantitative analysis, qualitative coding, ad effectiveness, segmentation — the actual stuff researchers do all day. With receipts to match.

I’m also going to feature work from other researchers in the field. We figure things out faster together than we do alone, and the conference circuit is not where the good stuff lives. The good stuff lives in DMs, Slack threads, and the moment a senior analyst says “wait, try this prompt instead.” I want to publish that.

What to expect

A few rundowns a month. Some will be tactical — the intro to Claude course I’m finalizing this quarter. Some will be opinionated — why synthetic respondents are the future. Some will be other researchers showing what they’re building. There will not be hype. There will not be performative thought leadership. There will not be a single “AI is going to revolutionize” sentence, because we both know revolution is mostly just spreadsheets.

If that sounds like your kind of newsletter, you can sign up at the top of this page. If it doesn’t, no hard feelings. There are forty other researchers writing about AI right now and one of them is probably a better fit.

---

A small thing you can use today

Since I told you this newsletter would actually be useful, here’s one quick thing.

A few years ago, I built a Slack bot that answers questions from a knowledge base. I trained it on a folder of research documents — past studies, methodology notes, common stakeholder questions — and dropped it into a Slack channel. Instead of researchers DM’ing me the same five questions every quarter (“what percentage of users do X,” “do we have data on X audience”), they tag the bot and get an answer in seconds. It paid for itself in saved hours within the first week.

If you want to build one yourself, this YouTube tutorial is the cleanest walkthrough I’ve seen. It uses OpenAI’s API, LangChain, and Slack — under 30 minutes start to finish, and you don’t need to be a developer to follow along. I built mine off this exact setup.

A few things to keep in mind if you’re a researcher building one:

• Don’t feed it raw participant data. OpenAI’s API isn’t HIPAA-compliant out of the box and the default settings can use your queries to improve their models. For anything with PII or sensitive client info, either anonymize aggressively before feeding it in, or swap OpenAI for a model that runs locally (Llama or Mistral via Ollama) or for Anthropic’s API with the Zero Data Retention add-on.
• The knowledge base is the whole game. If your underlying docs are messy, your bot will be confidently wrong. Spend the time cleaning your source material before you build anything fancy on top of it.
• Connect it to where the docs already live. The tutorial uses a local text file, but the real win is pointing it at Confluence, Google Drive, or Notion so it stays current without you maintaining it.
• Set expectations with stakeholders. Tell your team it’s a fast lookup tool, not a replacement for a researcher. Otherwise people will start asking it strategy questions and getting confidently wrong answers.

That’s the rundown. More soon.