NVivo AI Cloud: Advanced AURA Chat Prompts

Code suggestions and linking excerpts

Any change AURA wants to make to your project is displayed to you as a prompt. The prompt names what AURA is asking to do, and you decide with Reject or Approve; some prompts also offer Edit so you can adjust the suggestions before approving.

The suggestions cover the range of project changes AURA can propose, each with its own prompt, for example, adding codes to the codebook (AURA suggests adding these codes to your codebook), applying an existing code to excerpts, and linking to excerpts.

Saving answers to a memo

There are two ways an AURA answer becomes a memo. AURA can propose one when it judges an insight worth keeping, shown as AURA suggests saving this as a memo with the proposed title and body; use Edit to revise the draft (a memo is your first-person artifact), then Approve to save it or Reject to discard it. On approval, the memo appears in Memos.

Separately, you can save any AURA message directly with the Save to memo option on that message; this is your own one-click action and needs no approval step.

Finding exact mentions of a word or phrase

When your question is wording-sensitive ("find every mention of 'population'", "where does anyone say 'fishing'?"), AURA runs a literal, case-insensitive search over your source text. AURA chooses between the two search types automatically: an exact term or phrase goes to literal search; a meaning-based question goes to semantic retrieval.

AURA returns verbatim snippets, each sliced exactly around one occurrence and quoted as-is in quotation marks, attributed to its source and showing the linked cases and overlapping codes. Alongside the snippets it states coverage, for example "7 occurrences of population across 3 sources.".

Building a matrix or crosstab

Ask AURA a tallying question ("show me codes by Role as a grid", "count excerpts for each code by case") and it returns a dense table with separately computed row totals, column totals, and a grand total. Every cell is computed live from your data.

You can count by excerpts (the default), by sources, or by cases, and you can cross one dimension against another as long as the two differ. Dimensions can be codes, cases, sources, or a case attribute (named as it appears in your case list, for example Role).

Measuring how much of your data is coded

Ask a "how much is coded" question ("what percentage of Project's material is coded at each code?") and AURA returns coverage: the fraction of text covered, computed per code, per source, or per case. Coverage is always one-dimensional, so there is no crosstab here.

Finding the most common words

Ask for a word list ("what are the most common words used in the code - real estate pressure?") and AURA returns the top terms by frequency.

Additional Useful Chat Workflows

Citations

When AURA quotes evidence, it adds a citation that opens the exact passage in the original source. Use the citation to review the surrounding text and confirm that the claim is supported by the source, rather than relying on the summary alone. 

Suggested replies

After a reply, AURA may offer a short row of suggested replies, quick follow-ups you can click to continue the conversation without typing. They are generated for the specific prompt and refreshed when you reopen the AURA Chat.