AVA's knowledge has clear boundaries. Understanding those boundaries makes her more useful and helps you catch the rare case where her answer needs verification.

What AVA knows with certainty

AVA knows with certainty what is in your data lake at the moment you ask. Record counts, field values, revision history, relationships between records — all of this comes from a live query against your actual data. There is no uncertainty in these answers, only the ordinary possibility that the data itself is inaccurate (because someone entered it incorrectly in the source platform).

When AVA returns a query result, it is the actual answer from your actual data. If it looks wrong, the most likely explanation is that the source data reflects something unexpected — not that AVA made an error in retrieval.

What AVA does not know

Platforms you have not connected. If you have not connected Salesforce, AVA has no knowledge of your Salesforce data. She will tell you this explicitly rather than guessing.

Data added to platforms between sync cycles. For platforms without webhook support, ThatApp polls on a regular schedule. There can be a window of a few minutes between a change in the source platform and that change appearing in your lake. AVA will indicate when she is aware that a sync is in progress.

What will happen in the future. AVA can identify trends and patterns in your historical data. She will not make predictions she cannot support with your actual data.

Why something happened. AVA can tell you that a record's status changed on a specific date and was changed by a specific user. She cannot tell you the reason behind that decision — that lives in human judgment, not in the data.

How AVA signals uncertainty

When AVA is uncertain about something, she says so directly. Phrases like "based on the data in your lake," "I don't see a record matching that description," and "I'm not able to confirm that without seeing the connected platform" are her way of being precise about what she does and does not have access to.

AVA will not invent a record to fill a gap. She will not estimate a count when she can query the real one. She will not pretend to know something about a platform she cannot see.

If you ask AVA a question she cannot answer from your data, she will tell you what she would need to answer it — usually, connecting the relevant platform.

When to verify AVA's answer

Verification is worth doing when:

  • The answer is going to drive an important decision and you want to be certain the source data is accurate
  • AVA indicates that a sync was recently completed or is in progress
  • The question involves data from a platform you connected recently and the initial sync may not be complete
  • You have made significant changes in a source platform within the last few minutes

Verification is not usually necessary for routine queries against well-established platforms with stable sync history. AVA's answers in those cases are as accurate as the source data itself.

How to check AVA's work

For any query result, you can ask AVA to show her reasoning:

"How did you arrive at that count?"

She will describe the query she ran, including the filters and conditions she applied. This lets you verify that she is looking at the right data and applying the right logic.

For record-level results, you can cross-reference directly in the source platform. AVA can show you the source record URL for any result.

Related: How AVA Learns Your Organizational Data · How Your Data Lake Works · How to Query Your Data With AVA