> ## Documentation Index
> Fetch the complete documentation index at: https://nmem.guoxudong.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Ask for an evidence brief
> Turn saved context and a source into a concise analysis that distinguishes evidence from gaps.
AI Now · Lesson 3 / 5 \~5 min Intermediate
Question
Source
Evidence
Investigate
Save
Use the context and source in your Nowledge Mem AI Now task to compare the case for a trial with the reasons to hold back. Keep every finding traceable.
## What you'll accomplish
You'll get an evidence brief that separates supported findings from open questions.
## Learn with the simulation
On desktop, the simulation opens automatically when you enter this lesson. Follow the popup hints to send a request, read its reply, and check the result. If you close it, click **Start the simulation** under [Your turn](#your-turn) to reopen it. Each lesson prepares its fictional Atlas materials; no model setup is needed.
The simulation runs locally with preset replies. It does not call AI or search the web, and refreshing resets the practice. On mobile, read the steps first and use a desktop browser for the interactive exercise.
## Before practising in the real app
Continue the same conversation after verifying that AI Now can read the source from lesson 2. Keep the original document open for comparison.
## Your turn
Follow the hints in the automatically opened simulation. After practising, use the steps below to repeat the workflow with your own material in Nowledge Mem.
Ask for a brief that makes its reasoning reviewable:
```text theme={null}
Based only on the Atlas planning memories and atlas-planning-interviews.md, write an evidence brief about holding one cross-team planning review next quarter to check dependencies over the next twelve months. Leave future review frequency undecided. Do not search the web.
Include:
1. Findings that support a trial
2. Findings that limit or challenge it
3. Open questions, marked as internal questions or public research questions
4. Assumptions, kept separate from evidence
5. One small next step that would reduce uncertainty
For each finding, give the source name and section, what it establishes, and what it does not establish. Do not force a finding into a category if there is no supporting evidence.
```
Check one finding in the original source. For the practice material, these are useful boundaries:
| Source | Supports | Does not establish |
| ----------- | ----------------------------------------------- | ---------------------------------------- |
| Interview A | A need to spot dependencies earlier | That one cross-team review will solve it |
| Interview B | A trial using existing notes | Acceptance of more reporting work |
| Interview C | A gap: no trial or measured outcome is recorded | That the review works or fails |
Ask AI Now to revise claims that go beyond those boundaries. Missing measurement is a gap, not evidence against the review. The three interviews are fictional practice evidence, not a market sample.
Pick one uncertainty for lesson 4. "How much preparation can our team support?" needs an internal answer. "What documented methods help teams find cross-team dependencies before committing plans?" can be researched publicly. A lack of evidence is a valid finding.
Follow the popup hints; reopen the simulation if you closed it.
Lesson complete. Continue to the next lesson, or try it in Nowledge Mem.
## What you should see
The brief separates a reason to test the idea from proof that it works. Each finding has a source, and the next uncertainty has a clear way to investigate it.
## Why this matters
An evidence brief is more useful than a polished answer with hidden assumptions. It gives you a clear boundary between the knowledge you own and the information you still need.
## Go deeper
Read [AI Now](https://mem.nowledge.co/docs/ai-now) for file analysis and grounded task examples.