Drafting a Legal FAQ and Contract Glossary: Building Internal Knowledge with Generative AI

Hello, I'm Noriaki Asato, Representative Attorney at LegalAgent.

Consultations that reach the legal team from within the company often repeat the same themes. Questions such as "What is an NDA for?" or "I heard that service agreements come in contracts for work and quasi-mandates; what is the difference?" A service agreement is not a single contract type under the law; depending on the actual arrangement, it may be a contract for work, a quasi-mandate, or a mixture of both. Answering such inquiries from scratch each time eats into the legal team's limited time.

This is where glossaries and FAQs (documents that compile frequently asked questions and answers) are useful. If they are shared in advance, business units can resolve initial questions on their own, and the legal team can focus on individual consultations that require judgment. That said, writing them from a blank page takes effort. Generative AI (text generation services such as ChatGPT, Claude and Gemini) proves its worth in quickly putting together first drafts of such materials.

To operate an FAQ reliably within the company, it is necessary to make clear not only the wording of the answers but also the contract types covered, the intended readers and whom to consult. After asking generative AI for a first draft, the person in charge checks whether it fits the company's rules. Also decide who will be responsible for updates after publication.

Choose from Questions That Come in Repeatedly

Questions that make you feel "I've been asked this before" are strong candidates for FAQ items. If you give the AI past consultation records and anticipated questions and have it arrange them in Q&A format, the skeleton takes shape faster than writing by hand.

In legal work, themes that lend themselves to an FAQ include the following. Start by choosing a few questions from consultations you have actually received.

• Basic points to check when concluding an NDA (non-disclosure agreement)
• Points for the client and the service provider to note in service agreements
• Criteria for choosing between seals and electronic signatures
• Criteria for agreeing to the terms of use of cloud services (SaaS)
• Points to note when providing data that includes personal information to outside parties

Add Practical Points to Note to the Meaning of Terms

A glossary becomes a guide that business unit staff can use on the ground if, rather than simply listing dictionary meanings of terms, it also states "why it matters in practice" and "in what transaction situations it appears".

For example, when taking up "liquidated damages" (Article 420 of the Civil Code), organize it as follows. First, as the definition, show that it is an agreement that sets out in the contract in advance the amount of compensation to be paid if a default occurs. Next, as its practical meaning, explain that if the conditions for applying a valid liquidated damages clause are met, the predetermined amount can be claimed without individually proving the actual amount of damage, while for the paying side it can become an excessive burden, so it is important to check the reasonableness of the amount and the validity of the clause. Then, as situations where it appears, list penalty clauses in service agreements, sales contracts and the like.

If staff who read the explanation can consult the legal team by asking, "Is this an amount we could bear if we were the one paying?", the glossary leads to individual judgments. A stated predetermined amount does not mean it can be claimed for any kind of breach.

Intended Readers and the Scope of Answers

It is important for internal materials to use language that readers who are not legal experts can understand. If, when giving instructions, you specify "avoid technical terms, and when you use them, add a plain explanation", the text will be easier to use in practice.

For an FAQ aimed at business units, instruct the AI to encourage consultation when in doubt. The number and length of answers can be changed to suit the questions handled.

Please create a first draft of an FAQ on the following theme for business unit staff within the company. About 5 questions in question-and-answer format, please.

- Answers should avoid technical terms and be 3–5 lines. When using a technical term, always add a one-phrase explanation
- Where judgment can differ, write "This depends on the case, so please consult the legal team"
- Where stating something definitively would be risky (amounts or periods that change with circumstances, etc.), add a note saying so [Theme: e.g. "Basics of concluding a non-disclosure agreement (NDA)"]

For a glossary, ask the AI to add where each term is used in practice alongside its meaning. Check after the output whether the specific examples also apply to your company's transactions.

Please create a first draft of an internal glossary for the following legal terms. For each term, cover the following three points: (1) definition (in one sentence, without technical terms), (2) why it matters in practice, (3) in what contracts or situations it appears (one specific example).

Assume it will be read by business unit staff. For terms whose meaning changes with context, add a note saying so. [Terms: NDA, service agreement, acceptance inspection, etc.]

The following is an example answer about NDAs. Rather than having staff decide whether to sign based on this alone, add the company's approval procedure and whom to consult.

Q. A business partner has asked us to sign an NDA. Can I sign it right away? A. An NDA (non-disclosure agreement) is a contract that sets out the purpose of use and the method of managing the confidential information disclosed. Before signing, please check three points: (1) whether the scope of what must be protected (confidential information) is too broad, (2) how long the obligations last, and (3) the liability in case of breach. If you are unsure, please consult the legal team.

Including a note at the end of the answer to "consult the legal team" makes it easier for staff who are unsure to remember whom to consult. To make it even more effective, it is important to also make clear the rules for the point of contact, such as what materials and conditions to gather and whom to consult.

Checks before Sharing and the Person Responsible for Updates

If a general explanation created by AI is published as an internal document as it is, it may conflict with the company's business procedures and past operating policies and end up causing confusion on the ground. I think the appropriate division of roles is that the AI output is only a first draft, and people add and refine the company's own rules and decision criteria.

Generative AI can also produce plausible but inaccurate information (hallucinations). In particular, names of laws, article numbers and specific figures should be carefully checked against the original texts and official information. It is also essential not to state things uniformly and definitively, and to properly retain exceptions and reservations that depend on circumstances.

FAQs and glossaries are continuously referenced as the company's knowledge assets. Clearly state the date of creation, the person responsible for updates and the version of the internal rules covered, and put in place a structure for reviewing them regularly in line with legal amendments and changes to business procedures. Also, when entering past inquiry cases, check that they do not contain individual confidential information or personal information, and use them only after confirming that the environment is one the company has approved for use.