Data assistant

Pro This feature is available from the Pro edition.

The data assistant is an AI. It answers questions about your own data in your own words, such as “Which servers with a high protection requirement have no responsible person?”. It uses the language model set up for the AI assistant.

Data go to the model's provider

Unlike the help assistant, the model receives results from your inventory here: up to 25 assets with fixed fields such as name, category, location and health, counts, software figures and open deviations. It never receives notes, documents, credentials or settings. Use a model you trust with these data, such as your own model in your own network.

How an answer is made

1. The model chooses up to three read-only queries from a fixed list: find assets, count assets, find software, open deviations of check rules.
2. SentryAssets checks each query strictly and runs it with the asker’s rights.
3. The model answers only from the results. The results are listed below the answer with their filters, numbers and a link to the asset list.

Results can start no further query. Answers with links or hidden characters are not shown.

Switching it on

1. Set up the AI assistant first.
2. The administration switches the data assistant on separately on the page Data assistant, after confirming the sign-in.
3. Optional: names of responsible persons go along only with a second switch.

• The consent holds for the model’s endpoint at that moment. If the endpoint changes, it receives nothing until agreed again.
• Neither question nor answer is stored.
• Administration, information security, editors and auditors may ask; at most 20 questions per person and hour.
• The assistant has no writing queries and never changes data.

AI answers can be wrong; the listed results are the basis. Support, not an audit.