# Answer a research question (/use-cases/research/information-gathering)



Copy this prompt to start your first task:

<Prompt>
  Research [the research question and its intended use]. Break it into sub-questions, cite sources for every claim, and flag where sources disagree instead of blending them. Give me a clear, decision-ready conclusion, and check back daily for what's new if this keeps evolving.
</Prompt>

<LivePreview component="NewThread" scenario="{ draft: &#x22;Research [the research question and its intended use]. Break it into sub-questions, cite sources for every claim, and flag where sources disagree instead of blending them. Give me a clear, decision-ready conclusion, and check back daily for what's new if this keeps evolving.&#x22; }" />

## What is this workflow? [#what-is-this-workflow]

Answering a research question means giving a teammate a specific, scoped question instead of a vague topic, and letting it check multiple sources in parallel rather than one at a time. It merges repeated content across those sources, and flags conflicting reports instead of silently blending them into a single answer. The result is an answer where every claim links back to where it came from, so you can check the reasoning instead of just trusting it.

## How to use this workflow [#how-to-use-this-workflow]

<Steps>
  <Step>
    ### Hire a research teammate [#hire-a-research-teammate]

    Open **AI Teammates** in the sidebar, click **Hire AI Teammate**, and pick the **Researcher** template. Confirm the name and click **Create teammate**. They land in their own channel, which is where you'll send everything below. [More on hiring →](/ai-teammates/hire-an-ai-teammate)

    <LivePreview component="CreateAssistantPage" />
  </Step>

  <Step>
    ### Ask a specific question, with scope [#ask-a-specific-question-with-scope]

    A vague ask produces a vague answer. Give the scope, time window, and depth you need:

    <Prompt>
      You're a research-analysis AI, skilled at breaking a vague question into verifiable sub-questions and forming decision-ready conclusions from credible sources. Set up a "daily evidence update" automation for me for this research question: [the research question and its intended use]. On day one, break down the sub-questions, define credible-source standards, and find and cite sources; output a conclusion, the evidence chain, differing viewpoints, counterexamples, uncertainties, how I should use this conclusion, and what still needs verifying next. After that, add new evidence, new sources, and any changes to the conclusion each day — only report what would change my judgment or my next action. Prefer public sources; proactively offer to connect official databases and docs when a direct connection is available; and guide me through the browser when login is required.
    </Prompt>

    For example:

    <LivePreview
      component="AssistantChat"
      scenario="{ assistantName: &#x22;Researcher&#x22;, messages: [
    { author: &#x22;user&#x22;, text: &#x22;What are the current approaches to real-time collaborative text editing (CRDT vs. OT)? Focus on what's shipped in the last 12 months, not research papers. I need enough depth to make an implementation decision, with sources for every claim.&#x22; },
    { author: &#x22;assistant&#x22;, text: &#x22;Focusing on what's actually shipped in the last 12 months, not the research papers. I'll go deep enough to support an implementation decision and cite a source for every claim so you can check my work.&#x22; },
  ] }"
    />
  </Step>

  <Step>
    ### For an evolving topic, track it instead of asking once [#for-an-evolving-topic-track-it-instead-of-asking-once]

    Open **Automations** in the sidebar, click **Create via Chat**, and pick this teammate in the New Thread composer that opens:

    <Prompt>
      This is still developing. Check on it weekly, organize new developments into a timeline, and only tell me what's new since the last check, not the whole history again.
    </Prompt>

    <LivePreview component="NewThread" scenario="{ populated: false }" />
  </Step>
</Steps>

**Nothing here needs your approval:** searching, reading sources, and drafting the answer are all routine read actions, which run on their own. See [Control](/ai-teammates/control) for what does pause for sign-off.

## Tips for Better Results [#tips-for-better-results]

| Tip                                                  | Description                                                                                                                        |
| ---------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| Give scope, time window, and depth up front          | A bare topic gets a generic answer. Naming what to focus on (and what to exclude) gets a usable one the first time.                |
| Ask for a source on every claim                      | Lets you check the reasoning yourself instead of just trusting the answer.                                                         |
| Expect conflicts to be flagged, not resolved for you | Contradictory reports are surfaced, not silently blended into one answer — read those flags before treating the answer as settled. |
| Track evolving topics instead of re-asking           | Set up a weekly check via Automations and get only what's new since last time, not the whole history repeated.                     |

## Common Questions [#common-questions]

<Accordions type="single">
  <Accordion title="What happens when sources disagree?">
    The conflict is flagged rather than silently blended into one answer, so you can see where reports diverge.
  </Accordion>

  <Accordion title="Does it check sources one at a time?">
    No. Multiple sources are checked in parallel, and repeated content across them is merged.
  </Accordion>

  <Accordion title="Does every claim in the answer have a source?">
    Yes. The answer is built so every claim links back to where it came from.
  </Accordion>

  <Accordion title="Can I track an ongoing topic instead of asking once?">
    Yes. Set it up through Automations to check weekly and report only what's new since the last check, not the whole history again.
  </Accordion>

  <Accordion title="Does this need my approval to run?">
    No. Searching, reading, and drafting an answer are routine read actions that run on their own.
  </Accordion>
</Accordions>

<Cards>
  <Card title="Research" href="/use-cases/research" description="Back to the research use case overview." />

  <Card title="Watch competitors weekly" href="/use-cases/research/competitive-analysis" description="Track named competitors instead of a general question." />
</Cards>