Drafting Legal Documents in the Age of GenAI

Just over a year ago, a Head of Knowledge at a global law firm casually asked me, “Should we even bother with document automation anymore, Catherine, when Generative AI is taking over?”

It was an offhand remark, but it stuck with me. Since then, I’ve spent much of the past year trying to answer that question.

I’ve also heard variations of it in almost every conversation about document drafting. Some believe GenAI is the future, while others remain skeptical about its reliability for legal work. The debate is ongoing, and the landscape is evolving rapidly.

Breaking Down Legal Drafting: Tasks and Document Types

To cut through the noise, let’s break down document drafting into its core components. Rather than viewing it as a single process, I’ve found it helpful to categorise drafting into three key tasks and five main document types:

TASKS

• Creating a first draft
• Adapting or bespoking a draft
• Negotiating documents

TYPES

• Transactional (e.g. contracts, corporate documents)
• Litigation (e.g. witness statements, court forms)
• Advisory (e.g. legal opinions, client memos)
• Recording (e.g. board minutes, meeting notes)
• Reporting (e.g. lease reports, due diligence summaries)

Creating a First Draft: What Works Best?

Now, let’s explore the best tools for creating a first draft across these five document types.

1. Transactional Documents

The gold standard for creating the first draft of a transactional document isn’t GenAI, it’s document automation.

• Why? Document automation tools use deterministic if-then-else rules to ensure accuracy and consistency, while also significantly speeding up drafting.
• If your firm has an agreed precedent, automation is the best option.
• If not, an automated version of an industry-standard automated precedent (e.g., from FromCounsel or Practical Law) is the next best thing.

If no precedent is available, lawyers often work from a previous example. GenAI tools can be useful here, particularly for extracting key terms from an old document and adapting it for a new matter.

One law firm I worked with wanted to convert old leases (often received as PDFs) into new ones more efficiently. A legal GenAI tool was able to generate a new lease from the old one while highlighting every change for the lawyer’s review. Formatting is a crucial factor here, if a tool can’t maintain correct formatting, the lawyer ends up spending just as much time fixing the document.

Finally, in rare cases where no precedent or previous example exists, GenAI can help overcome blank-page paralysis. While early vendors and commentators suggested GenAI could draft contracts from scratch, in reality, lawyers rarely do this anyway! That said, GenAI can assist in brainstorming clauses and structuring agreements to get you that good starter for ten.

2. Litigation Documents

For court forms, automation is the best choice, it ensures compliance, precision, and efficiency.

For witness statements, the process is different. Lawyers typically draft these from notes or recordings of client interviews. GenAI excels at generating text from structured input, making it an excellent assistant for this task. Some tools even flag inconsistencies or missing details, which can be invaluable.

Interestingly, some jurisdictions have banned the use of GenAI in drafting witness statements, which I personally find unnecessary. In the U.S., some AI-powered tools can generate first drafts of defences, citing relevant cases and legal arguments. We’ve all seen the stories of AI hallucinations, so these vendors focus heavily on verification of citations.

3. Advisory Documents

A common complaint about written legal advice? “Too long, didn’t read.”

GenAI is fantastic at summarising and clarity improvements. It can also act as a sparring partner, e.g. lawyers can ask:

• “What have I missed?”
• “What questions would my client ask about this?”

This is an area where traditional document automation doesn’t add much value, but GenAI can be a game-changer.

4. Reporting Documents

For documents like lease reports, GenAI’s summarisation abilities might seem like a perfect fit. However, legal agreements often contain cross-references and interdependencies with one clause limiting another in a way that AI might not fully grasp.

In some cases, a structured Q&A approach (using deterministic automation) is safer and more reliable. Lawyers can better interpret complex contractual relationships than GenAI currently can.

5. Recording Documents (e.g. Board Minutes)

Anyone who has ever tried to automate board minutes knows what a pain they can be! The complexity of formatting attendees, chairpersons, and the sheer number of possible resolutions, makes them difficult to standardise.

Here, GenAI is incredibly useful. Feeding it meeting notes and an example format allows it to generate draft minutes quickly and accurately, saving significant time compared to manual drafting or traditional automation.

Bespoking and Negotiating Drafts

Once a first draft is created, the next task is bespoking, adding provisions not covered by the precedent and reviewing the document before sending it to the other side.

Options for improving this process include:

• Expanding precedents to cover more scenarios
• Using clause bank tools to build a library of frequently used provisions
• Leveraging GenAI tools to suggest relevant clauses based on context

When it comes to negotiation, GenAI-powered tools are already proving super useful for pre-execution document review.

• These tools compare a received draft against industry standards or a firm’s playbook, suggesting necessary amendments.
• Lawyers can then review the AI’s suggestions, accepting, rejecting, or modifying them as needed.

To get the best results, curating well-structured playbooks is more important than ever. A well-organised checklist of negotiation points ensures AI-driven reviews align with firm or client policies.

The Future: Combining GenAI and Rule-Based Tools

In summary, there’s no single answer to the question of whether GenAI or automation is the best solution for legal drafting, it depends on the document type and specific task.

• Use deterministic rule-based tools for drafting where accuracy is paramount, such as transactional precedents or court forms.
• Leverage GenAI for drafting witness statements, advisory documents, board minutes, and summaries, tasks where generating text from structured input is valuable.
• Expect hybrid solutions to become the norm. We’re already seeing GenAI-assisted document automation, where AI helps both build and refine automated templates and completes the questionnaires instead of a human to produce the document automatically.

Ultimately, I believe the focus will shift from whether a tool is GenAI-powered or not, to whether it enables faster, more consistent, and higher-quality drafting.

I’d love to hear your thoughts, what’s your experience with GenAI in legal drafting? Let’s continue the conversation!

Also, I wonder how out of date this article will be in 6 months time!?!? I have never known things to move so fast. If you are experimenting, or want to, DM me!