At the beginning of Star Trek IV: The Voyage Home—you know, the one with the whales (stay with me here)—the newly reborn Spock is completing a battery of tests administered by a Vulcan educational computer. The machine challenges him across disciplines: mathematics, astrophysics, history, logic, technical specifications. It demands synthesis, presses him to reason under pressure, and moves fast enough to keep a half-Vulcan genius working to keep up. When Spock gets an answer wrong, the computer does not move on. It pushes. It is, in every sense that matters, an intellectual adversary.

When the scene aired in 1986, a computer like that was pure fiction. Forty years later, it is available to everyone. With carefully curated source materials, AI can now do what the Vulcan computer did: test your knowledge, challenge your reasoning, expose the gaps in your analysis, and force you to think with more rigor than you would manage on your own. We have reached the point, in other words, where the educational computer from a science fiction film is a reasonable description of a tool you can build on a Tuesday afternoon.

For attorneys, this changes what is possible when no colleague is available, when the preparation window is short, or when the work is the kind that benefits from another mind in the room but has never justified the cost of putting one there.

A Different Use of AI

Most of the conversation about AI in legal practice centers on a single model: AI produces a deliverable, and the attorney reviews it. A draft brief. A contract markup. A research memo. The machine does the work. The human checks it. This is valuable, and I have written elsewhere about when it is and is not appropriate. But it is not the only way AI can make an attorney more effective, and it may not be the most important one.

There is a second model, less discussed, in which AI does not produce the attorney’s work product, but instead engages with the attorney’s thinking before, during, or alongside the work. The output of this second model is a better-prepared attorney.

The distinction matters. In the first model, AI substitutes for cognitive effort. In the second, it intensifies cognitive effort. The attorney thinks harder and more clearly because something is pushing back—asking the uncomfortable question, identifying the gap in the argument, refusing to let a vague assertion stand unchallenged. This is what a good colleague does when you walk into her office and say “let me run something by you.” The difference is that the AI version does not require the colleague to be available, interested, or sufficiently versed in the subject matter to engage.

The key ingredient is curation. A general-purpose AI making general-purpose objections is a parlor trick. An AI loaded with the actual record in your case becomes a different tool entirely. The quality of the intellectual partnership is a direct function of the materials you give it to work with.

Preparing for the Panel

An attorney preparing for oral argument on a contested motion faces a familiar problem. She knows her arguments. She has written the brief. What will decide the argument is the one thing she cannot reliably predict on her own: how those arguments hold up under pressure from a skeptical panel.

The traditional preparation is a moot: a senior partner or experienced colleague sits her down and plays the judges. This works well when it happens. It often does not happen, because the people qualified to run a good moot are the same people with no time to spare. The attorney ends up rehearsing in her own head, which is better than nothing but worse than facing real resistance. She walks into the courtroom having tested her arguments only against her own expectations of what the hard questions will be.

Seed an AI agent with the full motion file—the briefs from both sides, the relevant case law, any prior hearing transcripts from the matter—and instruct it to challenge the attorney’s position. The agent becomes a sustained, informed adversary, capable of pressing the attorney with arguments derived from the actual materials in the case. It finds the weak points in the argument because they are weak, not because it has a litigator’s intuition. The result is not a replacement for the senior partner’s moot, but it is available at eleven o’clock on Sunday night, when the argument is Monday morning and the senior partner is asleep.

The attorney who has spent two hours defending her position against an informed adversary—even an artificial one—walks into the courtroom having already faced the hardest version of the opposing argument. She has rehearsed answers to questions she had not anticipated. She has identified the concession she may need to make and prepared the pivot that follows it. None of that preparation shows up in a document. It shows up in the courtroom.

Deposing the Expert

The same structure applies in a different litigation context. An attorney preparing to depose an opposing party’s expert witness faces a version of the same challenge: she needs to understand the expert’s methodology well enough to probe its weaknesses, and she needs to anticipate how the expert will defend it.

Load the agent with the expert’s report, the expert’s prior publications or testimony if available, and the relevant technical standards. Instruct it to defend the expert’s position. The agent cannot replicate the expert’s personality or evasive habits; those human variables stay beyond its reach. It can, however, identify the strongest defenses the expert is likely to offer and force the attorney to develop lines of questioning that account for them. If the attorney’s planned cross-examination has a gap—a place where the expert can credibly deflect—the agent will find it, because the source materials contain the information the expert will rely on.

The attorney walks into the deposition room having stress-tested her examination against the expert’s own materials. She has a plan for the expert’s best answers, not only his worst.

The Socratic Mode

Not every intellectual partnership is adversarial. Some of the most productive engagements are closer to an interview than a cross-examination.

Consider the task of developing a business plan—whether for a new practice area, a client pitch, or an internal strategic initiative. The challenge is rarely a lack of ideas. It is the gap between having ideas and organizing them into a coherent argument with supporting specifics. Most attorneys can talk about their plan with more nuance and precision than they can write about it on first attempt because conversation forces a kind of real-time accountability that a blank page does not. Someone asks “what’s your revenue model?” and you have to answer, and in answering you discover whether you actually know.

An AI agent loaded with relevant market context and instructed to interview you about your assumptions, competitive positioning, resource requirements, and timeline produces that conversational accountability on demand. The agent draws out the specifics that you would otherwise leave vague. It identifies the places where your plan relies on an unstated assumption you have not examined. And it does this iteratively: your answers feed the next round of questions, so the conversation sharpens as it goes. The business plan you write afterward is better because you were forced to defend it before committing a word to the page.

The same dynamic applies to tasks like annual self-evaluations, where the challenge is less about substance than about articulation and organization. An attorney knows, in a general sense, what she accomplished over the past year. Getting that knowledge out of her head and onto the page in a structured, specific, well-allocated format is a different problem. An AI agent that asks targeted questions about outcomes, obstacles, and contributions that might otherwise go unmentioned organizes the material and jogs memory. The right question at the right moment surfaces accomplishments the attorney had forgotten or dismissed as routine. It pushes past generalities and helps allocate substance across the categories that the evaluation form demands but that the attorney’s memory may not naturally organize around. And it can flag when the attorney’s language is striking the wrong tone—coming across as defensive when she means to be factual, or aggressive when the context calls for something measured. The deliverable improves because the content is more thorough and because the attorney had a reader before the readers who matter.

What Remains Beyond the Machine

These tools have limits, and the limits matter. An AI intellectual partner does not replace the senior partner’s thirty years of courtroom instinct and judgment. It does not replicate the colleague who knows your client well enough to say “that argument is technically right but your client will hate it.” It cannot evaluate whether your gut feeling about a witness is sound, or whether a deal term that looks aggressive on paper is commercially reasonable given the relationship between the parties.

The tools are also only as good as what you give them. An agent built on thin materials will produce thin challenges. An agent built on the wrong materials will produce confident, plausible, and misleading challenges—which is worse than no challenge at all. Curation decides whether the tool sharpens your thinking or distorts it.

Availability

The best preparation has always put your thinking under pressure—a senior partner grilling you before argument, a colleague poking holes in your deal across the table. That kind of pressure is worth having. Getting it is the hard part: the right colleagues are rarely free exactly when you need them. So most attorneys prepare alone for work that badly needs the pushback.

AI brings the pushback. The judgment and the instinct stay with you, but at eleven on Sunday night, AI is in the room, loaded with your record and willing to argue. The Vulcan computer made even Spock work to keep up, and it would not let a wrong answer stand. Now your computer can do the same.

This publication is intended for general information purposes only and does not and is not intended to constitute legal advice. The reader should consult with legal counsel to determine how laws or decisions discussed herein apply to the reader's specific circumstances.