---
title: "Prepare your documents for an AI assistant"
description: "A four-step playbook to get policies, procedures and product documents ready for an AI assistant, so its answers are current, sourced and limited to what each person may see."
url: https://lidvizion.ai/guides/prepare-your-documents-for-an-ai-assistant/
site: Lid Vizion
index: https://lidvizion.ai/llms.txt
---

[Field guides](https://lidvizion.ai/guides/) 4 steps

# Prepare your documents for an AI assistant

A four-step playbook to get policies, procedures and product documents ready for an AI assistant, so its answers are current, sourced and limited to what each person may see.

For: Operations and knowledge owners about to connect company documents to ChatGPT, Claude or a custom assistant.

An AI assistant is only as good as the documents behind it. Duplicated, outdated or contradictory files produce confident wrong answers, and loose permissions expose information to the wrong people. These four steps clean up the sources for one use first, so the assistant starts accurate and stays that way.

## Step 1 — Pick the questions first

**Goal:** Limit the scope to what the assistant actually needs to answer.

1. List the twenty questions people ask most for the chosen team or task.
2. Note which documents currently answer each one.
3. Mark questions with no good written answer — those need writing, not connecting.
4. Leave everything else out of the first version.

**You finish with:** A question list mapped to source documents and gaps.

## Step 2 — Clean the sources

**Goal:** Make sure there is one current answer for each question.

1. Remove or archive duplicate and outdated versions.
2. Resolve contradictions between documents with the owner.
3. Add a clear title, owner and review date to each document.
4. Write short answers for the gaps found in step 1.

**You finish with:** A clean, owned set of documents for the first use.

## Step 3 — Set access and connect

**Goal:** Make sure people only get answers from documents they may see.

1. Group documents by who may read them.
2. Connect only those groups to the assistant, with matching permissions.
3. Exclude personal, legal and financial records unless the use requires them and access is controlled.
4. Test with accounts from different roles.

**You finish with:** A connected source set with access that matches your rules.

## Step 4 — Test and keep current

**Goal:** Check the answers and stop them drifting.

1. Run the question list and check each answer against its source.
2. Fix the document, not the prompt, when an answer is wrong.
3. Give each document owner a review reminder.
4. Add new sources only when a real question needs them.

**You finish with:** A tested assistant and a routine that keeps sources current.

## Common pitfalls

- Connecting an entire drive on day one.
- Fixing wrong answers with prompts instead of fixing the source.
- Documents with no owner or review date.
- Ignoring permissions until something leaks.

Want a hand getting started?

## Start with a workflow audit

We pick the workflow with you, set the baseline and write the bounded design.

[See the audit](https://lidvizion.ai/workflow-audit/) [Talk this playbook through](https://lidvizion.ai/contact/?need=ai-workflow&from=%2Fguides%2Fprepare-your-documents-for-an-ai-assistant%2F#write)

## See it as a workflow

- [Internal knowledge assistant](https://lidvizion.ai/use-cases/internal-knowledge-assistant/): Operations Answer staff questions from approved policies, procedures and product documents, with the source shown and a clear handoff when the answer is not there.

## More guides like this

- [Connect your CRM to the tools around it](https://lidvizion.ai/guides/connect-your-crm-to-the-tools-around-it/)
- [Build a vendor contract register](https://lidvizion.ai/guides/build-a-vendor-contract-register/)

## Related answers

- [How should a company prepare its data before using AI?](https://lidvizion.ai/answers/how-to-prepare-company-data-for-ai/)
- [What is retrieval-augmented generation, and when does a business need it?](https://lidvizion.ai/answers/what-is-retrieval-augmented-generation/)

Bring the messy version

## One workflow. A better way to run it.

Show us the tools, the handoffs, and the part that keeps slowing your team down.

[Let’s map your workflow](https://lidvizion.ai/contact/) · [Read our approach](https://lidvizion.ai/approach/)