Use AI to Draft Better Documents
AI Usecase Series
August 26, 2025
Large language models (LLMs) like ChatGPT and Claude are great for a quick second opinion, brilliant issue-spotter, and detail-oriented critic. Crafting an effective "prompt" (your request or task for the LLM) saves time and provides considerably better results; just remember to prioritize privacy and confidentiality.
Below is an example prompt to get you started, with workflow tips and a sample AI response.

LLMs have massive pools of comparative examples and expert resources (including the internet) from which to provide near-instant feedback. They're brilliant, tireless, and enthusiastically ready to help:
- Spot risks or ambiguities
- Identify missing concepts or sections
- Propose creative solutions
- Suggest better phrasing
Before you agonize over whether to ping a coworker for review, or send to a supervisor or client.
Example: AI-analying a Marketing Agreement
The Workflow:
Use CamoText to redact sensitive data in the draft text and anything else you want to provide → attach or paste the output text in the AI interface (chatgpt.com, claude.ai, etc.) with your prompt→ review the response, refine your draft, and repeat until you're satisfied!
Let's walk through a step-by-step example digital marketing consulting agreement.
Step 1: Protect Private Information
Load or paste your draft using CamoText, click Anonymize, and confirm that every term you deem sensitive has been replaced by an encrypted placeholder (search or highlight-to-anonymize anything additional you'd like to redact). For example, a company name might become <ORGANIZATION_a56cD0ac>
and a client name, <PERSON_5cA441f1>
.
No algorithm or AI is perfect: review the output for anything you deem sensitive.
The output text is what to use in your prompt. AI recognizes and distinguishes between placeholders without needing to know the original terms.
Step 2: Craft an Effective Prompt
Clear, structured prompts lead to better results and avoid future backtracking, so invest time in a strong initial prompt.
A basic formula to remember:
Role + Task + Instructions + Context
Role: You are a ________ for a _______
Task: Review, analyze, critique, and suggest improvements to the pasted/attached draft ________
Instructions: Given the draft and context provided, (1) note commercially-standard provisions for this type of document that are missing or should be improved as a priority, (2) spot any liability or compliance issues to be considered, (3) suggest improvements in clarity, and (4) discuss any strategic or practical considerations arising from this agreement that are notable, whether or not expressly in the draft.
Context: The draft ____ is pasted/attached, and you are assisting the ___ side of the agreement. This is a relatively low-value contract for digital marketing services. The parties have no other pre-existing agreements, and the deliverables should be precisely scoped. Please ask any necessary clarifying questions before you begin.
Step 3: Review the AI's Response and Incorporate
Expect a detailed response nearly immediately after submitting your prompt, including:
- Suggested edits for clarity
- Sections you may have overlooked, like indemnification provisions or IP ownership
- Creative options like milestone-based payments or clearer deliverable definitions
- Clarifying questions to help refine your draft
Rinse and repeat – don't be afraid to tweak your prompt and submit multiple times to see what different outputs you get; prompt ChatGPT and Claude simultaneously and then ask them to critique the other's response; try asking LLMs to create a better prompt for you to then redirect back to it!
Use these as a guide when revising your draft, not a final answer — you remain the responsible drafter. AI can hallucinate, doesn't have your full context, and can prioritize satisfying you over getting fully "correct" answers.

AI bots are eager to help!
Other Types of Documents
This anonymize → prompt → review and refine → finalize workflow applies to just about any type of document. More subjective or strategic drafts can benefit from more human-like context, and more technical or factual drafts can benefit from more specific instructions and contextual links.
Objective and factual drafts can benefit from specifying example sources or attaching relevant information ("ensure you review the attached data files in your analysis"). More subjective, strategic, and creative drafts can benefit from more human-like setting contexts ("the counterparty often starts from a fairly aggressive negotiating position" or "I'd like to come up with a new solution here, don't be afraid to think outside the box"). If you're using AI as an initial pass before sending to your lawyer, it can also be useful to include in your prompt what provisions or issues to highlight for further review.
Over time, you might save prompt templates for different usecases you frequently encounter.
Best Practices
- Data privacy and confidentiality are paramount: assume everything that goes into an LLM is being logged or collected. Removing sensitive terms should be the first step in your workflow, on your device (asking a third party LLM to do this defeats the purpose!).
- Structured prompts with clear intent lead to more actionable outputs: Your prompt should include Role + Task + Instructions + Context.
- These tools are collaborative, not replacements: AI provides a fresh, objective perspective that can uncover issues you might miss, but also can be overly eager to please and may hallucinate.
- Save your best prompts: over time, you'll build a library of usecase-specific that makes this workflow even faster and more powerful. We'll provide more examples in our next usecase series article!
The goal is not to replace your expertise but to amplify it. AI tools can make you a faster and more savvy writer and reviewer.