> ## Documentation index
> Fetch the complete documentation index at: [llms.txt](../llms.txt)
> Use this file to discover all available pages before exploring further.

# Ask a question in Analyze

How Sleuth turns a question in your own words into a plan, datasets, and an answer you can check.

[Ask in Analyze](https://app.sleuth.studio)

Analyze is where you work with Sleuth on a question. Each question becomes a project: the chat in the middle, the datasets beside it, and the results as Sleuth builds them.

## What happens when you ask

Sleuth reads your message and decides what kind of request it is:

| You ask | Sleuth |
| --- | --- |
| A quick question, like "What phase is zanidatamab in for biliary tract cancer?" | Answers in the chat, with sources. |
| A question that needs a dataset, like "Rank every bispecific in Phase 2 or later for solid tumors on differentiation" | Proposes a plan and the datasets it will build, and waits for you to approve them. [Plans and approvals](plans-approvals.md) |
| A question about data already in the project | Answers from your datasets, without rebuilding them. |
| A change, like "Add a column for each asset's lead indication" | Proposes the change to the dataset and makes it once you approve. |

If your question is unclear, Sleuth asks before it starts. Answer in the chat and it carries on.

## Write a question Sleuth can run with

- **Name the decision.** "We need to pick two partners to approach at BIO" gives Sleuth more to work with than "find partners."
- **Set the scope.** A modality, a target, an indication, a stage, a region, or a list of companies.
- **Say what counts.** If unmet need matters more than market size to you, say so. Sleuth builds those criteria into the columns.
- **Add what you have.** Upload a deck, a shortlist, or a spreadsheet, and Sleuth uses it. [Use your own documents and data](own-data.md)

## While it works

Sleuth shows its reasoning as it goes, then fills each dataset. A project can hold several linked datasets, each at its own level of detail, such as one for programs and one for deals. You can leave; Sleuth emails you when it's done.

The chat panel has a few tools to keep long projects manageable: pins for results you want to keep in view, a Files tab for your uploads and outputs, a task panel that shows what's running, and suggested follow-ups you can click.

## Keep going in the same project

Ask follow-up questions, ask for a chart or a deck, add columns, or narrow the rows. Questions about results you already have don't use credits. If a dataset changes after an answer was built, the answer shows that it's based on older data, with a **Regenerate** button.

Sleuth also remembers what you tell it within a project, and preferences you set across projects, so you don't have to repeat your scope or your criteria.

## Related

- [Plans and approvals](plans-approvals.md): Sleuth shows you what it will build before it builds it. Here is what you approve and how to skip approvals.
- [Datasets: rows, columns, and cells](datasets.md): What a Sleuth dataset is, which cells use credits, and how to change one.
- [Outputs and exports](outputs.md): What Sleuth hands back, and how to get each one out of Sleuth.

*Updated November 9, 2026.*
