> ## Documentation Index
> Fetch the complete documentation index at: https://help.guardbook.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# The onboarding questionnaire

> The eighteen questions, what each section is for, and why the answers matter more than they look.

Eighteen questions, one per screen, across four sections. Every answer feeds something later, so this is not a form to click through.

Each question has a **Why we ask** panel giving the reason and the clause behind it. Open it when a question seems odd. It usually is not.

## The four sections

> [accordion] 1. Business Context (4 questions)
>
> Your sector, headcount, services and regulator.
>
> Headcount means everyone, not just surveyors, because the clause 3.1 and 3.2 training duties reach everyone who handles data or uses AI. Services shape which client-facing documents need an AI disclosure under clause 4.3.

> [accordion] 2. AI Usage (5 questions)
>
> Which tools you use, what for, how many staff use them, whether client data goes through them and what kind.
>
> This is the section that does the most work. See below.

> [accordion] 3. Risk & Governance (7 questions)
>
> Whether you have a policy, who is accountable for AI, how often you review, whether you have done a risk assessment, the state of staff training, your risk appetite, and whether your PI cover extends to AI.
>
> Two of these deserve care: the accountable person needs to be a **name**, not a department, because clause 3.2 requires it. And the PI question feeds a line clause 4.3 requires your client terms to carry.

> [accordion] 4. Outputs & Gaps (2 questions)
>
> Whether you tell clients about AI use, and your biggest concerns.
>
> The concerns question is the only one that is not driven by a clause. It seeds your risk register and tells us what to show you first.

## The question that matters most

**"Which AI tools does your firm use?"**

Every tool you tick becomes a row in your Tool Register immediately. Miss one here and it is missing from your register, your policy and your client disclosures until someone adds it by hand.

> [!NOTE]
>
> Include everything staff use for work, even occasionally, and even if nobody thinks of it as an AI tool. Transcription, meeting notes, the assistant built into your email, search tools that now answer rather than list.
>
> There is a free-text option for anything not listed. Use it.

## Answer honestly, not aspirationally

Several questions ask about things most firms have not done yet: no policy, no risk assessment, no formal training, never told a client. Those are the normal answers, and each one is offered as an option rather than buried.

> [!WARNING]
>
> A questionnaire answered as you would like the firm to be produces a generated policy describing a firm that does not exist. You will then have signed a document committing you to controls you do not operate.
>
> If in doubt, answer down. The gap is the thing guardbook is for.

## Changing your answers later

Your answers are saved as you go, and the questionnaire can be picked up where you left it.

They also keep working after onboarding: your AI policy and client disclosure statement are generated from them, so a change to how your firm operates is a reason to revisit the answers and regenerate.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.