The organization
A small team runs a recurring conference. Each cycle, an open call brings in a stack of speaker applications, and the organizers — already juggling venue, program, and logistics — have to review every one carefully and fairly.
The problem
Reviewing an application well means more than reading it. For each one, an organizer researched the speaker’s background and proposed topic, checked the event’s published participation requirements, and wrote it all up for the team. That cycle repeated for every application and landed on the same few people.
What we built
An n8n workflow that handles the first research pass and builds a shared workbook, so organizers start with a consistent, source-linked picture of every submission. It’s an AI-with-tool-use build: the AI runs web searches to gather background on each speaker and topic, going beyond the application text. It’s in pilot, and the team checks the research against the cited sources.
How it works
- The organizer uploads the application export and defines the research questions and event context.
- The workflow creates a Google Sheet and writes in the raw submissions, unchanged.
- For each submission, AI research passes run on the speaker and topic using reliable public sources.
- The findings are organized into consistent columns with source links and explicit uncertainty.
- A per-submission Google Doc captures the written findings for anyone wanting more detail than the workbook provides.
- When the run finishes, the organizer gets the links by email.
What it produces
- A sortable research workbook, one row per submission, with consistent fields and source links
- A written summary document per submission
- Email notifications with links when processing completes
Where humans stay in the loop
The AI compiles source-linked research; organizers make the decision. The research is not a speaker recommendation or selection decision. Organizers verify relevant findings, decide whether a topic is a good fit for the event and audience, and compare the assistant’s research with what they would have found on their own.
Why it works
- First-pass research is repetitive. The repeated lookup steps are a good fit for automation.
- Clear questions narrow the AI’s role. With the research questions and event context defined up front, the AI gathers specific facts instead of making open-ended judgments.
- A shared workbook reduces compilation work. Organizers still review the sources and decide.
What to copy
- Write your research questions down before automating. Ask for public work and experience relevant to the event, reliable sources, and explicit uncertainty. Exclude protected traits, private information, rumors, and irrelevant personal beliefs or associations.
- Treat AI research findings as leads to verify — especially any finding that could affect a real person’s application.
- Produce both a comparison view and per-item depth. A sheet for sorting and a document for reading serve different moments in the review.
The applicant intake prioritization case study applies the same pattern to a different intake problem. New to AI with tool use? What is an AI agent? covers the basics.
