Questionnaires
Customers and auditors send security questionnaires: spreadsheets with with dozens or hundreds of questions about how you run security. In QuantAssure you upload the file and an AI agent drafts an answer to each question from what your organisation has already recorded, with a confidence rating and the sources it used. You then review the answers before anything leaves the building.
Open Compliance → Questionnaires in the sidebar. The page is described as "Upload security questionnaires for AI-powered auto-fill".
Uploading a Questionnaire
- Choose Compliance → Questionnaires → Upload Questionnaire.
- Give it a Name, such as "CAIQ v4" or "SIG Lite 2024". The name is how you find it later.
- Choose the File. The supported formats are XLSX, CSV and TXT, up to 10MB. Older
.xlsfiles are also accepted. - Choose Upload & Analyse. While the file uploads the button reads "Uploading...", and you are taken to the new questionnaire's page.
Spreadsheets (XLSX) and CSV files are read as text, so the questions are easy for the agent to pick out. PDF upload isn't supported yet: a PDF is refused with the message "PDF upload isn't supported yet; upload a .xlsx, .xls, .csv or .txt file." Copy the questions into a spreadsheet and upload that instead. Other file types, such as .docx, are refused the same way.
Owners, admins and members can upload. Viewers can open and read questionnaires but do not see the Upload Questionnaire button.
Uploading a questionnaire also adds its text to your organisation's AI knowledge base, and every answer is kept there too, so the agent can reuse good answers on later questionnaires.
The Questionnaires List
Each row shows the questionnaire's name, its file type (such as XLSX), the
date you uploaded it, a status pill and, once questions have been found, an
Answered meter reading, for example, 38/40.
The list opens newest first; use the sort menu to switch between
Newest and Oldest. With none uploaded the
page says "No questionnaires yet".
Statuses
| Status | What it means |
|---|---|
| Uploaded | Saved, and waiting for the AI agent to start. You will normally see this only for a moment. |
| Processing | The AI agent is reading the questionnaire and drafting answers. The page says "AI is analysing your questionnaire..." and that this typically takes 2-5 minutes depending on the number of questions. |
| Completed | The agent has submitted its answers and they are ready to review. |
| Failed | Auto-fill did not start or did not finish. The page shows a panel titled "Auto-fill didn't complete" with the reason. |
If auto-fill fails
The reason on the panel tells you what to do. The common ones are:
- Auto-fill hasn't been enabled for your organisation, or your organisation's knowledge base hasn't been set up. Contact support, then upload the questionnaire again.
- Another AI job is already running for your organisation. Upload the questionnaire again once that job has finished; you can watch it on Agent activity.
- The agent could not be started, or it finished without submitting answers. Upload the questionnaire again, or contact support if it keeps happening.
A failed questionnaire cannot be retried from its page. Upload the file again, as a new questionnaire.
Reviewing the Answers
Open a questionnaire to see its answers in question order. The header shows the file name, the status pill and the Answered meter (for example, "38/40 answered"). Each answer card shows:
- The question, numbered Q1, Q2 and so on, with its category (such as access control or encryption) when the agent could tell.
- A confidence pill: High when a policy or control directly supports the answer, Medium when related evidence needed some inference, and Low when there was little evidence behind it. A dash means no confidence was recorded.
- A review pill: Needs review until a person approves or edits the answer, then Reviewed.
- The drafted answer. It is usually two to four sentences, written to name the actual policies, frameworks and controls it relies on. Where your records don't show the answer, the agent is told to say that plainly with low confidence rather than guess.
- Source pills, such as "policy: Information Security Policy", showing what each answer was built from.
The answers are drafts. The AI only knows what is in your organisation's records, so a Low confidence answer is a prompt to check the facts, not a statement you can send. Read every answer before it goes to a customer or auditor.
Where answers come from
The agent searches your organisation's knowledge base for relevant policies, the status of your compliance controls and their assessments, evidence and audit records, and answers you gave to earlier questionnaires. Keeping policies and control statuses current is the best way to get High confidence answers.
Approving and editing answers
Above the answers the page reads "N of M answers reviewed. Only reviewed answers are public." Owners, admins and members can review; viewers see the answers without the buttons.
- Approve records that you read the answer and it is right as it stands.
- Edit opens the answer in a text box. Choose Save and approve to store your wording; editing counts as reviewing. An edited answer shows an Edited note, and its wording replaces the AI's.
- Approve all high-confidence (n) approves every unreviewed High confidence answer after you confirm. Medium and Low answers always need an individual look.
If the agent is run again on the same questionnaire, it leaves edited answers alone. A new AI answer replaces any other answer and clears its review, so it needs approving again. Approvals and edits are recorded in the audit log.
Publishing Answers on Your Trust Centre
A Completed questionnaire can be published as the "Security Questions & Answers" section of your public trust centre. Choose it under Settings → Trust centre → Published Q&A. Visitors see each reviewed question with its answer (the edited wording where there is one) and category, but not the confidence rating or sources. Answers that still say Needs review are never public. The Trust centre tab shows "N of M answers reviewed" beside each questionnaire.
If your organisation published a questionnaire before review was added, answers that a person had not edited are no longer public. Open the questionnaire and approve them to bring them back. Answers a person had already edited stay public.
Who Can Do What
| Role | Can |
|---|---|
| Owner, Admin, Member | Upload questionnaires, read them, approve and edit answers |
| Viewer | Read questionnaires and their answers |
Publishing to the trust centre is limited to owners and admins, because the Trust centre tab is.
How It Connects to the Rest of QuantAssure
- Policies and compliance: the answers are drawn from your policies and control statuses, so improving those improves the answers.
- Findings: questionnaires do not create findings.
- Audit log: uploads, completed auto-fills, and answer approvals and edits are recorded in the audit log.
- AI allowance: auto-fill is an AI job, so it counts towards your organisation's AI usage, and most AI jobs run one at a time for an organisation.
- Trust centre: completed questionnaires can be published, as above.