Use memory well

Recover from a missing or poor answer

An AI answer can be poor because the source is missing, outdated, outside your access or misunderstood. Your Mensara Mind holds the records the agent can use, while the model turns those records into a response. This guide helps you find which part failed and choose a useful correction.

Contents

Check the question’s scope

Name the exact person, organisation, event and time you meant. A fact about another customer or an earlier period may look relevant without answering your question. Ask the agent to separate the answer it can support from background or inference.

Ask what was read

Have the agent name the maps and source records it used. Open them. If it searched but did not open the results, ask it to read the promising sources before making a claim. A failed or partial search is different from a successful read finding no support. Finding the source explains the normal route.

Distinguish absent information from a boundary

If the record is not available, check the intended Mind, account and area. A refusal can reflect access; a current-view exclusion can reflect memory status. Neither is a reason to try another identity or copy protected information into chat. Zones and permissions explains what a missing result does not establish.

Check freshness and conflicting evidence

For today’s guidance, ask for current information. For a past statement, ask for the evidence from that period. If sources disagree and no decision resolves them, keep the uncertainty visible and ask an authorised person to decide.

Correct the part that is wrong

If the source is sound, point to the passage the answer misunderstood and ask the agent to explain the mismatch. If information is missing, save the missing source. If the saved record is wrong, propose one precise correction with the evidence behind it.

After a correction applies, open the record and ask the original question again. A better sentence in the conversation is not proof that the underlying source changed.

Fix the source or the explanation

Where did the answer fail?

First open the sources behind the answer and check their subject, date and scope. If the source is sound but the explanation is wrong, ask the agent to re-read and explain the mismatch. If the source is missing or wrong, supply the missing evidence or propose a correction. After any correction applies, read the record and ask the original question again.

Leave a precise unresolved gap

If the evidence still does not settle the question, say what is missing and who could supply it. Do not turn a lack of evidence into a confident answer. That gap is a useful result when it prevents someone acting on a guess.

Related

Ask what is current, then ask why · Save a note, meeting or useful fact · Correct saved information in your Mind

Practical guides for people and agents

Start with one useful job

Join the waitlist

Plain text for AI agents. Same content as the human page.

## For agents

- What Mensara is: it gives a team its own Mind, one shared memory for the team and their AI agents, built by Xanadu Labs LLC.
- Status: signup is not open. A person joins the waitlist at https://app.mensara.ai/#waitlist. Agents cannot create an account or sign up.
- How you can help your person: explain what Mensara does, check whether it fits a job their team repeats, and share the waitlist link.
- Agent references: https://app.mensara.ai/llms.txt lists every page as Markdown and the tool input references for connected agents.

---

# Recover from a missing or poor answer

Guides · Use memory well · [How it works](https://app.mensara.ai/how-it-works)

An AI answer can be poor because the source is missing, outdated, outside your access or misunderstood. Your Mensara Mind holds the records the agent can use, while the model turns those records into a response. This guide helps you find which part failed and choose a useful correction.

[Page as Markdown](https://app.mensara.ai/how-it-works/improve-an-answer.md)

## Check the question’s scope

Name the exact person, organisation, event and time you meant. A fact about another customer or an earlier period may look relevant without answering your question. Ask the agent to separate the answer it can support from background or inference.

## Ask what was read

Have the agent name the maps and source records it used. Open them. If it searched but did not open the results, ask it to read the promising sources before making a claim. A failed or partial search is different from a successful read finding no support. [Finding the source](https://app.mensara.ai/how-it-works/maps-and-navigation) explains the normal route.

## Distinguish absent information from a boundary

If the record is not available, check the intended Mind, account and area. A refusal can reflect access; a current-view exclusion can reflect memory status. Neither is a reason to try another identity or copy protected information into chat. [Zones and permissions](https://app.mensara.ai/how-it-works/zones-and-permissions#when-an-area-is-missing) explains what a missing result does not establish.

## Check freshness and conflicting evidence

For today’s guidance, ask for [current information](https://app.mensara.ai/how-it-works/current-and-evidence#ask-what-applies-now). For a past statement, ask for the evidence from that period. If sources disagree and no decision resolves them, keep the uncertainty visible and ask an authorised person to decide.

## Correct the part that is wrong

If the source is sound, point to the passage the answer misunderstood and ask the agent to explain the mismatch. If information is missing, [save the missing source](https://app.mensara.ai/how-it-works/save-information). If the saved record is wrong, [propose one precise correction](https://app.mensara.ai/how-it-works/suggest-a-correction) with the evidence behind it.

After a correction applies, open the record and ask the original question again. A better sentence in the conversation is not proof that the underlying source changed.

### Fix the source or the explanation

Where did the answer fail?

First open the sources behind the answer and check their subject, date and scope. If the source is sound but the explanation is wrong, ask the agent to re-read and explain the mismatch. If the source is missing or wrong, supply the missing evidence or propose a correction. After any correction applies, read the record and ask the original question again.

## Leave a precise unresolved gap

If the evidence still does not settle the question, say what is missing and who could supply it. Do not turn a lack of evidence into a confident answer. That gap is a useful result when it prevents someone acting on a guess.

## Related

- [Ask what is current, then ask why](https://app.mensara.ai/how-it-works/current-and-evidence)
- [Save a note, meeting or useful fact](https://app.mensara.ai/how-it-works/save-information)
- [Correct saved information in your Mind](https://app.mensara.ai/how-it-works/suggest-a-correction)

Previous page: [Save a note, meeting or useful fact](https://app.mensara.ai/how-it-works/save-information)

Next page: [Understand activity and the history of a change](https://app.mensara.ai/how-it-works/activity-history)

Activity records help explain which identity asked for something and what the Mind did.

[Join the waitlist](https://app.mensara.ai/#waitlist)