The team for this work
Which agents to hire, which Marketplace templates start them, and the order they need each other in.
✦ Guide · Intermediate
AI for research teams
Cited briefs, themes from interview transcripts, questions answered only from the attached sources and a weekly source sweep as a team of agents built around a citation rule.
Start guideYOUR OUTCOME
You will run a research pod where every claim is tied to a source it read, transcripts become themes with quotes, questions outside the sources are refused rather than filled, and new sources are filed weekly.
A claim without a source is marked, not smoothed over Themes from transcripts carry counts and quotes with their source The weekly sweep files new sources with a summary and a link
4 min
Hire the first wave
Import, rename, and give a goal with the deliverable, the sources and the finish line.
The group and the second wave
Run the team in one room; hire the rest once the first deliverable exists.
The routine that keeps working
One recurring job on a schedule, then a status round and a grade.
What each lesson covers
The guide, lesson by lesson.
The team for this work
Which agents to hire, which Marketplace templates start them, and the order they need each other in.
Start from the work, not the tool: Cited briefs, themes from interview transcripts, questions answered only from the attached sources and a weekly source sweep as a team of agents built around a citation rule. The team is Ren · Research (Research assistant), Ivy · Transcripts (Customer feedback analysis), Sources · Q&A (Document and policy Q&A), Noor · Findings (Presentation building). Each starts from a Marketplace template, imported as your own copy, renamed and re-goaled on the setup card; nothing personal and no creator credentials come with a template.
Practice. Open Marketplace, Agent templates, and read the Instructions and Integrations of each template on this list before you import anything. Note which sources each one will ask you for.
Marketplace → Agent templates → open Research assistant. Read Instructions, then Integrations.
Hire the first wave
Import, rename, and give a goal with the deliverable, the sources and the finish line.
Import agent creates your copy and opens the setup card. Rename it, replace the goal with yours, and choose When to work. The research assistant is the first hire, and the citation rule in its goal is the whole method: every claim tied to a source it read, unverifiable claims marked.
Practice. Import Research assistant, rename it Ren · Research, and paste your goal on the setup card. Choose Now, once, and Continue. Check its first deliverable in Doc Manager before the next hire.
Name: Ren · Research. Goal: Research the adoption of heat pumps in Australian homes since 2020 with sources. Tie every claim to a source you actually read, quote the figure as the source states it, and mark any claim you could not verify. Deliver a two-page brief with a numbered source list. When to work: Now, once.
The group and the second wave
Run the team in one room; hire the rest once the first deliverable exists.
Create group chat, name it Research pod, add Ren · Research, and post what the team is for. Then hire the second wave with goals that start by reading the first wave's document. @Name directs one member, @everyone the room, Work ↗ opens a member's own session. Transcripts become themes with counts and quotes; the sources agent refuses questions the papers do not answer; the findings deck traces every slide to the brief or the themes.
Practice. Create Research pod, add each hire as it joins, and ask every member to post what it made and the one thing it would flag.
@everyone Welcome to the research pod. Ren's brief is in the workspace. Ivy: turn the attached 24 interview transcripts into themes with counts and three quotes each, naming the transcript. Sources: answer the attached question list from the papers only. Noor: build the findings deck from Ren's brief and Ivy's themes.
The routine that keeps working
One recurring job on a schedule, then a status round and a grade.
New sources appear every week; a sweep files them so the brief can be kept current without re-reading the field. Every run is a session under Finished today. Close each cycle with a status round, "@everyone one line each", and a grade against the acceptance checks each agent wrote first.
Practice. Create the routine, use Run now to get today's first report, then run one status round in Research pod.
New routine. Name: Weekly source sweep. Instructions: Search for new reports, papers and datasets on the named topics published since the last run. For each, file a one-line summary and a link in Doc Manager, and mark anything whose publisher you could not identify. Frequency: Every week, Monday, 08:00.
Keep going
Make the learning stick
Start with one real piece of work.
Open Oppermind beside the lesson, apply each step, and leave with something you can use.