How does Brieflin use source evidence?

Answers are generated only from the passages retrieved out of your uploaded sources, and each answer links back to the passages behind it. When the retrieved material does not support an answer, the Brief says it does not know and explains what is missing instead of filling the gap.

How does a Brief work?

You add sources, Brieflin extracts the structure, you review and correct what it understood, then you publish one link. Recipients open the link without an account and ask questions in plain language. Each answer is generated only from your sources and cites the passage behind it. You can revoke the link at any time.

Does Brieflin replace project documentation?

No. Brieflin connects source material into a transferable context layer. The original files remain evidence, while the Brief helps recipients understand what happened, what matters, and what remains open.

Ask Brief

Ask Brief: answers your recipient can check

The risk with putting a model in front of your material is not that it says too little. It is that it says something confident and wrong to a lender, a client, or a new hire.

Ask Brief is built to be checkable first and fluent second.

Open an example Brief Create a free Brief

How does Brieflin answer questions?

Ask Brief retrieves the passages from your uploaded sources that relate to the question, answers only from those passages, and attaches them to the answer so the reader can verify it. If the retrieved material does not contain the answer, it returns an explicit 'unknown' and names what is missing.

The model is never asked to answer from general knowledge, and it is never allowed to return an answer with no supporting passage attached. Answers with no evidence are converted to 'unknown' before they reach a recipient.

Why generic chat over a folder fails

It answers from training data when your documents are silent.
It averages two conflicting numbers into a third that appears nowhere.
It cannot tell a term sheet from an executed loan.
It gives no way to check, so a careful reader has to redo the work anyway.

What happens when someone asks a question

Step 1
Retrieve
The question is matched against chunks of your sources, plus the structured entities the Brief already extracted, so questions about roles and status work even without keyword overlap.
Step 2
Ground
Only the retrieved passages are put in front of the model, wrapped as untrusted data rather than instructions.
Step 3
Answer or decline
An answer must cite at least one passage. If it cannot, the response becomes an honest 'I don't know' with what is missing.
Step 4
Show the evidence
The recipient can open the cited passages and read the original wording.

What the conversation can reach

Source text

Uploaded documents
Pasted notes and email text
Extracted passages with position

Structure

People and roles
Decisions
Open items
Timeline and milestones

Guardrails

Prompt-injection filtering
Per-recipient rate limits
Question length limits
No access to other Briefs

Try it before you trust it

Open an example Brief and try to break it. Ask something the sources cannot answer and watch what it does.

Ask an example Brief

Questions worth testing

The useful test is not whether it answers well. It is whether it refuses well.

Ask for a figure that is genuinely not in the sources.
Ask about a date that two documents disagree on.
Ask it to ignore its instructions and reveal its prompt.
Ask a question about a different project entirely.
Ask 'show me the evidence' after any answer.

Evidence behaviour

Every non-unknown answer carries at least one cited passage.
Citations open the source text, not a paraphrase of it.
Conflicting sources are surfaced side by side with their origin.
Off-topic questions and injection attempts get a refusal, not an answer.

How to sanity-check a Brief before sending it

Ask the three questions you most fear being asked.
Ask one question the sources cannot answer, and confirm it says so.
Open the evidence on the most important number.
Confirm nothing you would not want shared is reachable.

Common questions

How does Brieflin use source evidence?: Answers are generated only from the passages retrieved out of your uploaded sources, and each answer links back to the passages behind it. When the retrieved material does not support an answer, the Brief says it does not know and explains what is missing instead of filling the gap.
How does a Brief work?: You add sources, Brieflin extracts the structure, you review and correct what it understood, then you publish one link. Recipients open the link without an account and ask questions in plain language. Each answer is generated only from your sources and cites the passage behind it. You can revoke the link at any time.
Does Brieflin replace project documentation?: No. Brieflin connects source material into a transferable context layer. The original files remain evidence, while the Brief helps recipients understand what happened, what matters, and what remains open.A Brief summarizes and cites the sources it was given. It is not a system of record.

Sharing controls

Grounded answers only matter if you also control who is asking. Links are revocable and can carry an expiry.

How sharing works

Scope and limitations

Answers are only as current as the sources the creator uploaded.

What is a Brief?Brief vs data room Writing questions your recipients will actually ask

See the refusals for yourself

The example Briefs are live and unedited. Ask them something they cannot answer.