The AI Due Diligence Gap Legal Leaders Can’t Afford to Ignore

Someone you know is struggling with depression and asks an artificial intelligence (AI) platform how to end their life.

Instead of intervening, directing them to emergency resources or discouraging self-harm, the system validates their thinking and, as some reported cases suggest, provides information that may facilitate suicide.

That is a catastrophic failure. Now consider a different setting.

A lawyer asks an AI platform to analyze a regulation, investigate a witness, summarize evidence or research a legal issue.

The system invents citations, fabricates authorities, misstates facts and delivers flawed analysis with the serene confidence of someone who has never once been wrong and does not intend to start now.

But this failure can stay hidden until a client catches it, a regulator questions it, a judge exposes it or a reporter turns it into a social media spectacle.

One failure appears obvious. The other appears. Authoritative.

Everybody vets software. Almost nobody vets assumptions.

Legal leaders have spent years asking what AI can do. Competitors rushed to showcase innovation. Clients demanded updates and answers.

Far fewer have asked the question that matters most: As AI embeds itself in legal operations, how will clients, courts, regulators, carriers and the media judge those decisions when something goes wrong?

Things go sideways; they always do. The real question is whether firms will be ready when things go wrong.

That gap matters because transformative technologies rarely do the most damage through capability alone. The real damage comes from scrutiny, doubt and the collapse of confidence when expectations are not met.

What Clients Are Really Asking—and It’s Not What You Want to Hear

Clients are not lying awake wondering which large language model you chose or how elegant your implementation roadmap looks.

They want to know how confidential information is protected, how outputs are validated, who is in charge and who will answer for the failure when something goes wrong.

That reality exposes a blind spot in how many organizations evaluate AI. Dashboards can measure utilization rates, productivity gains, implementation milestones, cost reductions and adoption trends with impressive precision.

What they cannot measure is credibility. And credibility matters most when something fails.

Trust is not built by technology alone. It is earned through transparency, accountability and the confidence that someone is prepared to answer difficult questions when things do not go as planned.

Three Questions Legal Leaders Should Ask Before Their Clients Do

If firms are serious about AI governance—and serious about keeping client confidence—three questions demand immediate attention.

1. What are clients no longer willing to pay for?

Every industry eventually sees valuable capabilities become expected. Organizations do not always recognize when a differentiator has become an expectation.

The strategic risk is failing to see how quickly client expectations move.

The strategic question is no longer what AI can do. It is what clients will continue to pay for once capabilities that once justified premium rates become widely accessible.

As expertise becomes easier to access, its value increasingly depends on what surrounds it: judgment, context, commercial understanding and the ability to help clients make difficult decisions under uncertainty.

Action: Identify the five services clients value most today. Then ask a harder question: If those services became faster, cheaper and universally available tomorrow, what would still differentiate the firm?

2. What are we not seeing that our clients are? 

Every CEO, board and general counsel possesses information that never appears in legal memos, board materials, earnings calls or regulatory filings.

They know where decisions are really made, where priorities are shifting, where frustrations are building and where opportunities are emerging.

Understanding a client’s legal matters is not always the same as understanding the forces shaping the client’s business.

The question is not whether that information exists—it’s how much of it your firm is hearing in time to matter.

The most consequential changes affecting clients often occur long before they become legal issues. Firms that recognize those changes early gain a level of commercial understanding that legal expertise alone cannot provide.

Action: Ask your most important clients a simple question: What has become materially more important to your business during the last two years? Then listen carefully—not only to the answer, but also to what is absent from public discussions.

3. Which parts of the firm survive only because they already exist?

Every successful organization inherits assumptions so old that nobody remembers making them.

Over time, those assumptions become structures, compensation models, reporting lines, processes and accepted truths that survive largely because challenging them has become uncomfortable.

The danger is that success can camouflage obsolescence. Practices that once created real advantage can keep running long after the conditions that justified them have disappeared.

If a competitor launched tomorrow with no legacy systems, no historical assumptions and no obligation to protect the past, what would they build differently? The answer may reveal opportunities and vulnerabilities that years of success have obscured.

Action: Identify three aspects of the firm that would look fundamentally different if they were designed today rather than inherited from the past. Then ask why they have not changed. The answer will either validate the current model or reveal assumptions that no longer deserve protection.

What AI Cannot Commoditize

Every law firm asks how AI will change legal work. Far fewer ask how AI will change what clients need, value and expect. That is the harder question—and the one that will decide who wins.

Clients who can access expertise more easily, generate information more cheaply and solve routine problems more independently will not stop needing lawyers. But they may become more selective about where they seek advice, what they are willing to pay for and which relationships they value most.

The firms that thrive will not be the ones that adapt fastest to AI. They will be the ones that adapt fastest to clients adapting to AI.

About the authors: 

Richard Torrenzano is chief executive of The Torrenzano Group, which helps organizations take control of how they are perceived™. For nearly a decade, he was a member of the New York Stock Exchange Management (policy) and Executive (operations) committees. His new book is Command the Conversation: Next-Level Communications Techniques.

Ronald J. Levine is a litigation attorney with more than 45 years’ experience who lectures and writes on artificial intelligence. He co-chaired Herrick, Feinstein LLP’s litigation department and also served as its general counsel. He authored a series of law practice articles for major legal media that are used by attorneys throughout the country. He graduated Harvard Law School with honors and Princeton University summa cum laude.