It's not that AI is hard. The problem is that your law firm doesn't want to change anything

View on LinkedIn

I have been consulting for law firms that want to implement AI. After dozens of meetings, training sessions, and demos, the pattern is so predictable I could automate it by now. Everyone arrives with the same enthusiasm and leaves with the same excuses.

What is curious is that everyone knows AI is not the future — it is the present. They watch the impressive demos, feel that gnawing FOMO when their competition shows off its new tools, pay for expensive training programs, buy courses that promise to revolutionize their practice. And then, nothing. Silence. They go back to their desks and keep doing everything exactly the same way they did ten years ago.

Why? Because real efficiency is not on the surface, and nobody wants to do the hard work of digging. They want results without the process, transformation without discomfort, innovation without bothering anyone. It is like going to a nutritionist and asking for a diet that still lets you eat all the chocolate you want, drink all the wine without restriction, and — please — no exercise.

My father runs a notary's office.

It is the perfect example of this paradox. He sends me videos of gurus talking about ChatGPT with the excitement of someone who has just discovered fire. He does use AI, of course — but as an enhanced Google, a reference tool. Meanwhile, his transcriptions are still done manually, quotes take hours when they could be instant, and every instrument could have an automatic summary with an index for quick identification, but it does not.

I have tried to show him that a system could take the data entered by the client and generate an initial draft in his own structure and style, saving him literally half the time on operations that require drafting. His response is always the same: "the clients won't want that," or "that is part of process X," or my personal favorite,

"for such a small change, we're better off doing nothing."

And it is not just my father. It is every lawyer, every partner, every managing attorney I sit down with. The pattern is identical: they arrive wanting to automate EVERYTHING, preferably by tomorrow, with fantasies of an Amazon-style layoff. We start digging into their daily work, identify a perfect workflow to automate, and then the dance begins. "No, not that one — it's part of process X." Fine, let's refocus on data capture. "The clients won't want that." Alright, something smaller then. "For something that small, we're better off doing nothing."

Or they arrive expecting the magic product — that mystical ChatGPT that will solve their entire firm without them lifting a finger. Spoiler: it does not exist and it never will. Real automation is built through your conscious selection of tools and your sustained commitment to training. What it truly costs is not money. What it costs is mapping real processes, being willing to make radical changes, and requiring everyone to use the new tools without allowing them to do things "the old way." The first implementation will go wrong, the second will be mediocre, but the tenth will be perfect. The problem is that nobody wants to go through the first nine.

Meanwhile, more and more clients not only want you to use AI — they demand it. And they are right. They understand that not knowing how to use these tools in 2025 is a professional literacy problem, and they are not going to subsidize your fear or your ignorance with their fees. When a corporate client that already uses AI across all its processes sees that its outside counsel is still working as if it were 2010, it quickly calculates how much that inefficiency is costing them. And believe me, it is considering its options.

What real implementation looks like

Real implementation starts by reducing ONE process. The one that takes you an hour — bring it down to 20 minutes. That seems small, right? Forty minutes saved. But if you manage to do that with 10 different processes, and you run each one at least once a week, that is 347 hours recovered over the course of a year. That is nearly nine full weeks of work. Nine weeks your competition is using to close more cases, to train, to live. While you are still "studying" whether it is worth trying.

And that is just the time savings. Because every process you automate teaches you something no course can give you. You learn which types of tasks you can delegate entirely versus which ones require mandatory human review. You develop the judgment to know when an error stems from your own poorly structured instructions and when it comes from real limitations of the system.

You understand how to design checkpoints and validations that catch problems before they reach the client. You learn to create workflows where technology handles the repetitive with industrial consistency while you focus on what truly requires professional judgment. By the time your competition finishes "learning about" AI, you will have already iterated 50 times and will be operating in a completely different league.