Most in-house legal leaders I talk to have already found an AI tool they're testing. Often more than one. They've seen a good demo, they can picture it working, and they'll tell you plainly that the status quo isn't fine.
Yet they still haven't bought anything.
Why is that? For a while I thought it was that they hadn't pinned down the problem they were trying to solve. It wasn't. Knowing the problem you're solving matters, it's just good discipline, but it isn't the thing holding them back. What's holding them back is the fear of getting it wrong.
The real fear isn't missing out
There's a good book on this called The Jolt Effect. The authors studied thousands of sales conversations and found that most deals don't die because the buyer picks a competitor. They die because the buyer is afraid of making a mistake, so they choose the one option that feels safe: doing nothing.
Legal feels this harder than almost any other buyer, and for a good reason. A general counsel's job is to be the person who catches the problem before it becomes a problem. That instinct is why they're good at the job. It's also why "right 95% of the time" doesn't sound like a win to them. It sounds like a one-in-twenty chance that a wrong answer goes out with their name on it. When the downside of being wrong is your credibility, "mostly right" is not reassuring. It's the thing keeping you up at night.
So the fear isn't missing out. It's messing up. And a tool that's merely impressive does nothing to answer it.
Something real did shift
Here's the part that's changing. Brenda Hansen captured it well after Priori's FlexFest recently: the era of AI experimentation is ending, and the era of adoption has begun. She's right. And it changes the math. Once adoption is the baseline, waiting stops being neutral. It quietly becomes a decision, with a cost. The bar for what legal software should do already moved, and the expectation of vendors moved with it.
The trouble is what the market does with that truth. Nearly every message aimed at legal turns it into pressure. Don't fall behind. Own your AI stack. The teams that move now will win. Move fast.
I understand why vendors say it. Urgency sells. But think about who's on the other end of that message. You're telling a person whose entire professional reflex is risk-avoidance that the risk is moving too slowly. That lands as pressure, not urgency. And a pressured risk-avoider does the most defensible thing available to them, which is to wait for more certainty, run one more pilot, ask one more question. The message meant to speed them up is exactly what freezes them.
A better question
Brenda's other point is the one that actually unlocks this: human judgment stays irreplaceable. Hold onto that, because it changes what you're shopping for.
The question isn't "what's the most powerful legal AI tool I can buy." That question rewards the flashiest demo, and the flashiest demo is usually the one most likely to confidently tell you something wrong.
The question isn't which legal AI tool is most powerful. It's which one makes it hardest to get this wrong.
The better question is simpler: what makes it hardest for me to get this wrong?
That's a different tool. Once you're asking it, a lot of the demo goes quiet and a few plain things start to matter more than anything on the feature list:
- Does it only answer from what my team has already approved? Not the open internet, not a model's best guess. My policies, my guidance, my own approved knowledge base.
- Does it show its work? Every answer should point to the exact source it came from, so anyone can check it in ten seconds instead of trusting it blind.
- Does it know when to stop? The most dangerous thing an AI can do in legal is answer a question that actually needed a lawyer. A tool you can trust is one that recognizes the hard question, stops, and hands it to a human with the context attached.
- Does the team's judgment stick? When a lawyer corrects an answer, that correction should become the standard answer for everyone next time, not disappear back into a chat thread.
None of it is exciting, but all of it is what separates a tool that could embarrass you from one built so the safe choice and the useful choice end up being the same choice.
Just start
Nobody gets fired for waiting. That's precisely why waiting is the most expensive thing legal is doing with AI right now. Moving fast and moving carefully are not opposites here. The teams that will look smart in a year aren't the ones who bought the most powerful thing. They're the ones who found the tool that made a mistake hard to make, and then just started.
