In the AI Era, Contract Review Standards Are Best Organized as a Legal Playbook

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

Business units ask for an urgent review, and sales staff press for a quick reply to the client. The speed of contract review is a challenge corporate legal departments face every day. Legal staff do not merely check whether clauses are lawful; they decide on proposed revisions based on the background of the transaction, the balance of power with the counterparty, the history of internal decision-making, past negotiations and so on.

Using generative AI is an effective means of reducing the time spent on preparatory work for a review. Tasks such as detecting missing provisions, identifying general legal issues, preparing first drafts of revised wording and polishing the wording of comments can, in some cases, reduce the time a person would spend writing from scratch.

However, to speed up reviews, internal decision-making procedures should be reviewed along with introducing the tools. The underlying question is what risk tolerance standards the company has as an organization and whether they are shared in writing. In my view, setting out these standards explicitly as a "legal playbook" forms the foundation of corporate legal work that uses AI.

Identifying Why Decisions Stall

In many cases, the progress of contract review stalls not only because the clauses themselves are difficult, but because the company's internal decision criteria have not been settled.

Issues such as how high a cap on damages to accept, how many years confidentiality obligations should survive, whether to prohibit subcontracting outright or require prior consent, and whether to require ownership of intellectual property rights or settle for a license are not determined uniquely by the text of the law alone. The range of acceptable risk varies with the transaction amount, the attributes of the counterparty, the nature of the services provided and the company's bargaining power. It is natural that a standard SaaS subscription costing tens of thousands of yen a month and a large contract outsourcing core operations call for different depths of clause review.

When the decision criteria exist only in the head of a particular person, decisions depend on that person. The policy on revisions may shift each time the person in charge changes, the background may have to be explained again each time an outside attorney is engaged, or loading a contract into AI may produce only abstract generalities. Introducing AI without settled internal standards only speeds up routine checks, and decision-making still stalls at points that require risk judgments specific to the company. Only once the company's decision criteria are set out explicitly does it become possible to quickly verify whether the output AI presents is appropriate.

Writing Down the Company's Standards for Each Contract Type

For AI tools to function effectively in contract review, it is important, before adjusting prompts, to organize the company's internal decision criteria into a document that both people and AI can interpret.

As a first step, identify the clauses to check with priority for each contract type handled routinely. For an NDA (non-disclosure agreement), first check the definition of confidential information and the purpose of use. Also consider the scope of permitted disclosure to third parties, the treatment of residual information, the term and the right to seek an injunction.

For a service agreement, confirm whether it is a contract for work or a quasi-mandate, and set the deliverables and acceptance conditions. Consider whether subcontracting is permitted and the conditions for damages, and also check the ownership of intellectual property rights and the handling of personal information.

For SaaS terms of use, check the requirements for interruption and suspension, data use and the level of security. Consider the content of the limitation of liability, and also confirm the procedures for cancellation and changes to the terms.

For fundraising agreements, check the investors' prior consent matters, preemptive rights and information rights. Read the founders' obligations and share transfer restrictions, and also consider the impact on the next round.

For each issue identified, assign a decision level such as "mandatory revision," "decided individually depending on the transaction terms," "accepted as a rule" or "requires prior confirmation with the business unit." If the decision levels are made concrete, the company can respond with consistent standards even when the person in charge changes or work is outsourced.

For example, once the company's policy is decided, you can tell AI things like "as a company guideline, a cap on damages of up to about 12 months of the monthly fee is accepted as a rule," "in transactions involving personal information, a consent procedure for subcontracting is mandatory," and "output the draft reply for the counterparty and the memo for internal review separately." These are examples of internal policies, not uniform legal standards. Note that the accuracy and effectiveness of AI output are not guaranteed, and even where internal standards are set, final review and judgment must be carried out by human staff.

What to Record in the Playbook

When creating a legal playbook, there is no need to aim for an enormous manual from the outset. In addition to the clauses to check, record the reasons for decisions and whom to confirm with.

Start by writing the key issues for each contract type. For each type that frequently arises at your company, state which clauses should be reviewed with priority.

Next, record the company's risk tolerance standards. Set policies on the cap on damages and the contract term. Set standard decision lines for issues on which decisions tend to differ from matter to matter, such as the ownership of intellectual property rights and the confidentiality period.

Also decide the matters to confirm with the business unit. The appropriateness of the consideration, the actual specifications of the work, the flow of data in systems and the like cannot be determined without confirming the facts with the people on the ground. Listing the confirmation items in advance reduces needless back-and-forth between legal and the business unit.

Prepare internal comments and comments for the counterparty separately. Internal comments should set out the degree of risk, alternatives and advice that takes into account the balance of bargaining power, while comments for the client should present explanations that convey the purpose of the revision without causing friction, and acceptable proposed revisions.

Also keep a record of past negotiations and the reasons for decisions. Being able to refer to the decisions made in the past for similar counterparties and contracts of the same kind makes it easier to maintain consistent handling.

Working with Outside Attorneys and Steps for Creating a Playbook

Preparing a legal playbook also makes the work product more useful when using outside attorneys or legal outsourcing.

Outside attorneys can answer legal questions quickly, but if the company's business background and the boundaries of acceptable risk are not shared, their advice tends to be accurate as legal doctrine yet out of step with the realities of the business. For example, some companies have a policy of not allowing any analytical use of customer data, while others allow the use of statistically processed data that meets the requirements of the law. In development settings that assume the use of cloud services, the necessary scope of subcontracting needs to be confirmed.

If the company's decision criteria are already compiled as a playbook, outside attorneys can confirm the company's policy and then proceed with the review while asking for any missing background information.

When creating a playbook, a realistic starting point is to choose one contract type handled in large numbers, such as NDAs or standard service agreements, and look back over the review comments from roughly the last 10 to 20 matters. Extract the revisions that are repeated routinely, and for items on which decisions have differed, put the reasons (such as the size of the amount, or whether the company is the outsourcing party or the service provider) into words as conditional branches.

Check Items for Building the System and Operational Support

Here are check items for confirming whether your company's contract review standards are functioning at the organizational level.

• Are the contract types that frequently arise at the company identified?
• Are the clauses to check with priority defined for each contract type?
• Are the decision levels for mandatory revisions and individually decided items clearly distinguished?
• Does the company have standard criteria for damages, confidentiality, IP ownership, personal information, subcontracting and termination?
• Are the facts to be confirmed with the business units on the ground organized in easy-to-understand terms?
• Are internal review comments and revision comments for the counterparty clearly written separately?
• Are past review histories and the reasons for decisions managed in a form that can be referred to?
• Has the company distinguished between information that may be entered into AI and restrictions on confidential information?
• Are the person responsible for checking AI output and the final decision-maker clearly designated within the organization?
• Is the background information to be shared with outside attorneys standardized?

In contract review, the company's judgment about the risks it will take on is reflected in the clauses. At LegalAgent, through collaboration between generative AI and attorneys well versed in corporate legal work, we provide consistent support from contract review work to Legal Playbook Development Support and building a legal function designed with the introduction of AI in mind. Companies that wish to put their decision criteria into words step by step while carrying on with day-to-day review work can contact us through the service below.

• Legal Outsourcing