Ask a company what contract terms it accepts, and you may receive a playbook.
Ask why it accepts those terms, when it makes exceptions, how different provisions interact, and what would change the decision, and the answers often become less precise.
Much of an organization’s legal judgment has never been fully articulated. It lives in the heads of experienced lawyers who have learned, over hundreds of matters, what the company will accept, what deserves concern, and when an ordinary issue has become unusual.
Humans have been quietly compensating for this gap for years. AI will make it visible.
Consider a lawyer reviewing a supplier’s request to use company data. The written policy may prohibit the use of confidential information for unrelated purposes. But the lawyer’s actual analysis is more nuanced. What type of data is involved? Can it be deidentified? Will the supplier use it to improve a service available to competitors? What security controls apply? Does the company have enough leverage to demand a narrower position?
The lawyer may reach an answer quickly because prior negotiations, internal conversations, business priorities, and hard-earned pattern recognition all inform the decision. Very little of that reasoning may appear in the official policy.
An AI agent does not reliably inherit undocumented institutional knowledge. If the organization has not defined the relevant factors, the agent must either guess, apply the written rule too rigidly, or ask a human every time.
None of those outcomes is especially attractive.
This problem extends well beyond contracts. Companies often lack clear definitions of what is acceptable, exceptional, and prohibited. Policies use terms such as “material risk,” “reasonable protection,” and “appropriate approval” without specifying how those standards should be applied. Escalation paths depend on who happens to know whom. Exceptions become precedents without ever being recorded as such.
Experienced lawyers keep the system functioning because they know where the written rules are incomplete. They remember why a prior exception was granted and whether it should apply again. They understand that two individually acceptable provisions can create an unacceptable result when combined.
This is legal judgment in practice. It is contextual, relational, and often tacit.
The arrival of agentic systems creates pressure to make more of that judgment explicit. If an agent is expected to negotiate, approve, renew, or make commitments, it needs more than policies and contract text. It needs to know which facts matter, how they affect the analysis, what range of outcomes is permitted, and when the matter must move to a person.
The difficult work is not automating the judgment. It is articulating it first.
That process will reveal uncomfortable inconsistencies. Two experienced lawyers may apply the same policy differently. A stated zero-tolerance position may have dozens of approved exceptions. A company may discover that its “standard” reflects habit rather than a deliberate risk decision.
Those discoveries are useful. They show where the organization needs clearer policy, better data, or an honest acknowledgment that the decision remains inherently contextual.
In-house teams can begin by examining recurring decisions rather than trying to document every possible legal judgment. Choose a common issue and ask: What facts change the answer? What outcomes are routinely acceptable? What is never acceptable? Which exceptions have been approved, and why? When does the combination of otherwise ordinary terms require escalation?
The answers can support playbooks, approval rules, training, and eventually AI systems. They can also improve consistency among human lawyers today.
Not every aspect of legal judgment should be reduced to a rule. Some decisions genuinely require experience, interpretation, and accountability. Making judgment explicit does not mean pretending that every answer is predictable. It means identifying what the company knows, what it has decided, and where real discretion begins.
AI may become very good at applying legal judgment. First, it will expose how much of that judgment companies have never actually captured.
Olga V. Mack is the CEO of TermScout, where she builds legal systems that make contracts faster to understand, easier to operate, and more trustworthy in real business conditions. Her work focuses on how legal rules allocate power, manage risk, and shape decisions under uncertainty. A serial CEO and former General Counsel, Olga previously led a legal technology company through acquisition by LexisNexis. She teaches at Berkeley Law and is a Fellow at CodeX, the Stanford Center for Legal Informatics. She has authored several books on legal innovation and technology, delivered six TEDx talks, and her insights regularly appear in Forbes, Bloomberg Law, VentureBeat, TechCrunch, and Above the Law. Her work treats law as essential infrastructure, designed for how organizations actually operate.
