Thoughts on AI in the Law: A Powerful Tool, but Not the Right Tool for Every Job.

Thanks to legal document platforms and AI systems, anyone can generate a contract, form an LLC, or draft a will in a matter of minutes. And that’s not always a bad thing. These tools are convenient, relatively inexpensive, fast, and in many cases, genuinely helpful.

But there’s a crucial misunderstanding happening when someone uses one of these tools: People assume that generating a legal document is the same thing as practicing law. It isn’t. And the difference matters most when something goes wrong. And more importantly, people often place confidence in those documents as if they carry the same weight as advice from a qualified attorney.

A generated document can look professional, it can seem comprehensive, and it can even include a lot of the “right” language. But the real question is: Was this the right approach for this situation in the first place?

What AI and Form Platforms Do Well

AI and form tools provide a good solution to a real problem. They are built for speed and standardization. They take common, repeatable scenarios, and apply pre-built solutions for many day-to-day situations. For example, with the proper instructions, these tools can do a really good job at helping someone form a single-member LLC with no outside investors, or draft a basic independent contractor agreement, or help a company implement a simple confidentiality agreement. These scenarios involve relatively well-known legal issues, and the range of acceptable solutions is narrow. That’s where these tools shine, because in straightforward scenarios, a standardized solution can be good enough. But what about situations where good enough isn’t good enough?

These tools are designed to answer the question: “What does this type of document usually look like?” But rarely do these tools answer: “What should this deal look like in light of the specific risks, industry, people, and goals?”

For example, consider a scenario where two co-founders adopt a 50/50 operating agreement without addressing deadlock risk. If a dispute arises, the business could end for both co-owners. Or perhaps a company uses a template contract that omits meaningful liability policies and procedures for its specific industry. Without this liability protection, the company is exposed to otherwise avoidable, and substantial, risk. Or imagine an investor comes into a company, but control and economic rights are not properly aligned. This investor could contribute a substantial amount, but receive very little in return. Or, from the company’s perspective, relatively little could be contributed, but now the new investor has substantial control.

In each of these scenarios, the document exists…but the structure behind the document is incomplete.

The Mistake: Treating the Document as the Outcome

This is where many business owners unintentionally create risk and liability. They treat the document as the finish line, but the document is not the job. The document is merely evidence that the job has been done…but has the job been done well?

The real work happens before and around that document. It happens in decisions about who controls the business, how money flows, what happens if things go wrong, how to eliminate ambiguity, and how disputes will be resolved. If these types of decisions are not made intentionally, the document will reflect that lack of clarity, and could potentially compound the problem.

This leads to the more important question: “If the document isn’t the real work, what is?”

What a Lawyer Actually Does

Much of a lawyer’s work isn’t drafting documents—it’s designing outcomes that hold up under real-world conditions. In these settings, the difference between a document and a well-structured outcome becomes very clear. When a qualified lawyer is involved, he or she doesn’t merely draft language. Rather, the lawyer is someone who thinks through problems that are not obvious at first glance, and structures outcomes that hold up under real-world problems.

Many legal problems result from issues clients didn’t even know to look for. A founder may not realize that equal ownership can create deadlock. A business owner may not fully understand the exposure created by a personal guarantee, or not adhering to required formalities. A partnership may never discuss what happens if one party stops contributing. These risks are not visible in a template. They are discovered either through the company going through a bad experience (the less-desirable and more costly way), or through legal experience, expertise, and thoughtful planning (the preferred and less costly way).

Examples

Structuring the Deal

For example, before a single word is ever drafted into a legal document—like an operating agreement—the real work is deciding how the deal should function. Should ownership track capital? Should distributions be equal, preferred, or tiered? Should certain decisions require a majority vote, a super majority, or unanimous consent? These are structural decisions, and they are rarely as simple as they first appear.

“Ownership tracks capital” sounds fair—until one investor contributes services instead of cash. Equal distributions feel straightforward—until one party is taking on extra risk or fronting early capital. Requiring unanimous consent sounds protective—until a single minority owner can hold the entire business hostage over a nonmaterial issue.

These aren’t drafting issues. They are design choices that require someone to ask:

• What happens if things go well?

• What happens if things go poorly?

• Where are the pressure points in this relationship?

• Who has leverage—and who needs protection?

This is where most AI tools and form platforms quietly fail. They assume the structure is already correct. They take the inputs they’re given and produce a document that perfectly reflects those inputs—even if the inputs themselves are flawed.

A lawyer’s role is different. A lawyer doesn’t just document the deal. A lawyer helps build the deal—testing assumptions, identifying imbalances, and forcing clarity on issues that the parties often haven’t fully thought through. Because once the structure is set, the document will enforce it—whether it was a good idea or not. Many of the most expensive mistakes in a deal are rarely found in the wording of the document. They’re found in the structure the document was drafted to enforce.

For example, consider two partners starting a business together. One brings the money, the other brings the idea and runs the day-to-day operations. They agree to split ownership 50/50. It feels fair. It is fair—at least at the beginning. They use a popular template to generate an operating agreement. The document is clean, professional, and internally consistent. It reflects exactly what they agreed to: equal ownership, equal voting rights, and that major decisions require both partners’ approval.

For a while, everything works. The business grows, revenue comes in, and the relationship is strong. But a few years later, things change. The investor wants to sell, but the operator doesn’t. The operator wants to reinvest profits, while the investor wants distributions. Maybe a third party makes a serious acquisition offer—but it requires a quick decision. And now, every one of those decisions requires unanimous consent. There is no tie-breaker, no buyout mechanism, no clear path forward when the partners disagree. The agreement—the one that looked complete and fair—does exactly what it was designed to do: it requires both partners to agree. So nothing happens. The deal stalls, the opportunity passes, and tension builds. Eventually, lawyers get involved—but now they’re not structuring a deal. They’re trying to unwind one.

Nothing in the document was “wrong.” In fact, the document was doing its job perfectly.

The problem was the structure. The partners built a system that only works when they agree—and didn’t plan for what happens when they don’t. That’s not a drafting issue. It’s a design issue.

And it’s exactly the kind of issue that templates and AI tools are not built to catch—because they assume the underlying deal already makes sense. When people think about legal documents, they tend to focus on the wording. But many real-world disputes don’t come from unclear sentences. They come from clear agreements that produce bad outcomes under stress. Because once the structure is set, the document will enforce it—whether it was a good idea or not.

Let’s consider the same deal, but structure it differently. The same two partners start a business together, where one brings the money, and the other brings the idea and runs things day-to-day. They still agree to split ownership 50/50.

But this time, before anything is drafted, they spend time working through how the deal should function when things don’t go according to plan. They’re forced to ask different questions:

• What happens if one of us wants out and the other doesn’t?

• What happens if we disagree on whether to sell?

• What happens if one of us stops contributing at the same level?

The answers shape the structure. They still have equal ownership, and they still share in the upside. But they add:

• A defined buy-sell mechanism if one partner wants to exit.

• A tie-breaker structure for major decisions (using a neutral advisor or a rotating control mechanism).

• A clear framework for distributions vs. reinvestment.

• Protections that recognize the difference between capital at risk and operational control.

The operating agreement that follows looks just as clean and professional as the first one - maybe even cleaner. But it does something different. It doesn’t assume the partners will always agree. It anticipates that they won’t—and gives them a path forward when that happens. On the surface, the two deals look nearly identical: the same partners, the same ownership split, and the same type of document. But one is fragile, and the other is resilient. One works only as long as the relationship works, while the other is built to carry the relationship through stress.

When the investor later wants to sell and the operator doesn’t, there’s a process. When a third-party offer comes in, there’s a way to resolve the decision. When priorities diverge—as they almost always do—the agreement doesn’t freeze the business. It guides it—not perfectly, and not without friction. But forward.

The difference isn’t in the drafting. It’s in the thinking that happened before the drafting began.

Both agreements are well-written. Only one is well-designed. Good drafting makes a deal enforceable. Good structure makes it workable.

Allocating Risk Intentionally

Every agreement is a mechanism for allocating risk. Who bears the cost if something fails?What happens if obligations aren’t met? Is liability capped, and are there exceptions?

Those aren’t boilerplate questions. They determine who carries the downside when things don’t go as planned. And things don’t go as planned: a missed deadline, a defective product, a deal that doesn’t close, a third-party claim that no one anticipated…just to name a few. And at these critical events, no one is asking what the contract says in theory. They’re asking who is actually on the hook. This is where risk allocation becomes real.

For example, a contract might include a standard limitation of liability clause, stating that liability is capped at the amount paid under the agreement, and consequential or indirect damages are not available. If you have a $50,000 contract with a $50,000 liability cap, that sounds reasonable—until you ask:

• What if the failure causes downstream losses far exceeding the contract price, say $500,000?

• What if one party is relying heavily on the other’s performance to meet its own obligations?

• Should certain risks—like confidentiality breaches, IP misuse, or gross negligence—be treated differently?

These aren’t drafting tweaks. They are decisions about who absorbs which risks—and at what scale. This is where AI tools and templates can sometimes mislead their users. They can include a limitation of liability clause, and an indemnification provision. But they struggle (and sometimes fail) to evaluate whether:

• The cap is meaningful in light of the actual risk.

• The carve-outs align with how the parties operate.

• The indemnity is broad enough (or narrow enough) to function as intended.

• The combination of provisions creates gaps, overlaps, or unintended exposure.

A lawyer’s role is to address these issues deliberately. Not just to include the clause—but to calibrate it. Because in practice, risk allocation isn’t about what sounds balanced. It’s about what actually happens when something goes wrong.

Most contracts don’t eliminate risk. They decide where it lands.

Stress-Testing the Outcome

Strong legal work assumes that something will go wrong. Not because the parties expect failure—but because experience says that over time, circumstances change, incentives shift, and even good relationships get tested. So, the real question isn’t whether problems will arise. It’s whether the agreement is built to handle the problems when they do.

What if a partner wants out early? What if additional capital is needed and one party won’t contribute? What if there’s a disagreement about direction? These aren’t fringe issues. They’re real pressure points that happen all the time. And when they show up, the agreement doesn’t get interpreted in the abstract. It gets applied—often quickly, and often under stress. This is where many agreements fall short. They read well, cover the basics, and include standard provisions. But, they haven’t been pressure-tested. They don’t answer:

• How does someone actually exit?

• What happens, step-by-step, when a capital call is missed?

• Who has authority when consensus breaks down?

Instead, they defer to silence, ambiguity, or provisions that only work if everyone continues to cooperate. This is where default language—and AI-generated language in particular—tends to break down. It can describe common scenarios, and include standard remedies. But it doesn’t anticipate how those provisions interact under stress, or whether they produce workable outcomes in real life.

A lawyer’s role is to run those scenarios in advance. A good lawyer should ask:

• If this happens, what does the agreement actually do?

• Does it move the parties forward—or does it stall?

• Does it resolve the problem—or escalate it?

Because once a dispute arises, the goal isn’t to have a well-written document. It’s to have one that works. A contract isn’t tested when it’s signed. It’s tested when something goes wrong.

Drafting for a Hostile Audience

When a deal starts to break down, each side is looking for leverage—and reading the same contract in very different ways. That’s the environment a contract has to function in, which means drafting isn’t just about clarity. It’s about control. It requires eliminating ambiguity, closing gaps, and anticipating how language will be interpreted by someone who is motivated to read it in the most adversarial way possible. For example:

• A phrase like “reasonable efforts” may sound balanced—but what does it actually require? What’s reasonable under the circumstances? And how would someone unfamiliar with the parties and the deal (e.g., a judge) know?

• A timeline that says “as soon as practicable” may feel flexible—but how long is too long? Should the consideration be paid within a week? A month? Six months?

• A provision that seems clear in isolation may create confusion when read alongside another section.

These aren’t academic exercises. They’re the arguments that get made when a deal starts to unravel. This is where many documents—especially those generated from templates or AI—tend to fall short. They haven’t been drafted with a hostile reader in mind. They don’t account for:

• How a counterparty might exploit ambiguity.

• How provisions interact across the document.

• How a court or arbitrator might interpret competing readings.

A lawyer approaches drafting differently. Not just asking, “Does this make sense?” But asking, “How could this be challenged?” “How could this be misunderstood?” “What happens if someone tries to use this against us?” Because in a dispute, no one is reading the contract generously. They’re reading it strategically. And a document that reads well at the outset can still fail.

A contract isn’t judged by how it reads when everyone agrees. It’s judged by how it holds up when they don’t. Two contracts can say nearly the same thing on the surface. But one leaves room for argument, while the other purposely removes it.

Where AI Fits In

In light of everything that’s been written above, it might surprise you to hear that lawyers should be using AI. In fact, as these tools become more capable, choosing not to use them is less a philosophical stance and more a practical limitation—one that can negatively affect the quality and efficiency of a lawyer’s work. This article itself reflects that approach. AI was used as part of the drafting process—but the structure, analysis, and conclusions are solely the author’s.

When used properly, AI doesn’t replace legal work—it changes where the effort is spent. Less time on first-pass drafting, more time on structure, risk, and outcomes. In our practice, that shift looks like this:

• Using AI to generate an initial draft—so the real work can focus on how the deal is structured.

• Using it to surface missing provisions or potential gaps that need to be addressed.

• Running “what if” scenarios to stress-test how an agreement behaves under pressure.

• Comparing alternative approaches to a deal before locking into one path.

• Translating complex provisions into plain English to eliminate ambiguity so clients can actually understand what they’re agreeing to.

Used this way, AI raises the ceiling. It makes the work more efficient—but also more thorough. It accelerates the process—but also creates space for better thinking. And that’s the real advantage. Not that AI produces a document faster—but that it allows more time to be spent on the parts of the work that matter the most.

But it’s critical to understand where that advantage stops. AI can generate language, recognize patterns, and suggest what is typical. It can also produce work that appears polished, complete, and authoritative—even when it isn’t. That’s not a theoretical concern. For example, in a well-publicized case, attorneys submitting a brief with citations generated by AI that appeared legitimate, but didn’t actually exist. The case—Mata v. Avianca—highlighted a core issue: AI can produce information that is convincing, but not real.

More often, the problems are quieter—but just as detrimental. AI can provide confident answers that don’t account for jurisdiction-specific law, omit critical provisions simply because they weren’t requested, rely on outdated rules, and produce documents that look complete but contain internal inconsistencies.

Moreover, AI cannot—and should not—make judgment calls about:

• How a deal should be structured.

• How risk should be allocated.

• How provisions will operate in the real world.

• How a document will be interpreted when the relationship breaks down.

Those decisions depend on context, experience, and an understanding of how things actually play out over time. Only a skilled attorney, often with the assistance of AI, can competently:

• Understand your full situation beyond basic instruction provided.

• Ask the follow-up questions that uncover hidden risk.

• Weigh competing risks and choose between them.

• Account for nuance across jurisdictions and fact patterns.

• Take responsibility for the outcome.

In other words, AI can help produce the work, but it cannot decide what the work should be. AI makes it easier to draft a contract, but it doesn’t make it easier to get the deal right. The risk isn’t that AI produces bad work. It’s that it can produce work that looks good enough to trust—when it isn’t. Like any tool, its output must be verified, refined, and applied with judgment.

The Right Tool for the Right Job

AI has changed how legal work gets done, but it hasn’t changed what good legal work requires. AI and legal form platforms are not a replacement for legal work. They are a different category of tool. When used appropriately, and with an understanding of their real-world limitations, they can be efficient and cost-effective. But when they’re used as a substitute for legal judgment, real problems occur. That’s because a legal document is only as good as the thinking behind it.

AI can help generate language, surface issues, and refine a document. But it cannot replace judgment. The distinction is simple: A document answers what was written, but a lawyer answers what should have been written. That difference becomes more important as the stakes increase.

If your situation is simple and low-risk, a standardized solution may be entirely appropriate. But if your situation involves meaningful money, long-term relationships, investors, or transition planning, the question changes. Not: “Can I generate a document?” But: “Have I structured this correctly—and will it hold up when it matters most?”

The goal isn’t just to have a contract. It’s to have one that works when it matters. Used well, AI raises the standard. Used poorly, it can create the appearance of completeness without the substance behind it. The right structure creates momentum, while the wrong one unavoidably limits it.

This article is for general informational purposes only and does not constitute legal advice or create an attorney-client relationship.