AI Isn’t Replacing Lawyers. But It’s Supercharging Institutions

A letter arrives on a Tuesday — from a landlord, an insurer, a bank or a government office. It was written by a machine, in seconds, for almost nothing.

By the weekend it is still on the kitchen table, beside a yellow highlighter and a cup of coffee gone cold, while the person it is addressed to tries to work out what it is asking, what it is threatening and what they must do before the deadline buried in paragraph 6. The letter was made by software. The person reading it is still a person. That asymmetry is the whole story.

I spent more than two decades as a civil litigator in Ontario, working the cases that arrive in writing and do not go away when you close the envelope: divorces, denials, foreclosures — the worst weeks of ordinary people’s lives. I once walked through a client’s home and found duct tape running along the floor, across the kitchen island and two feet up the walls on either end, bisecting the room into his half and hers. People who cannot afford a lawyer to draw the line for them draw it themselves. Sometimes in ink. Sometimes in tape.

You have been told AI will replace lawyers. That is the wrong story. What AI is actually doing is upgrading the institution on the other side of you — giving the side that already had more lawyers, more time and more money the power to produce more pressure, faster, for almost nothing. The lawyer is not being replaced. The institution that hires the lawyer is being turned into something you cannot match.

Most legal disputes are not won on truth. They are won on endurance — by the side that can absorb one more letter, one more motion, one more month of suspended life. AI does not reverse that advantage. It multiplies it.

The gap was already wide. Statistics Canada has found that of Canadians facing a serious legal problem, only one in three contacts a legal professional. In family court, up to 80 per cent of people now appear without a lawyer. This number has been climbing for more than a decade.

Across the aisle, a 2025 global survey found that active use of generative AI by legal organizations had nearly doubled in a year, to 26 per cent from 14 per cent, with most firms expecting it to become central to their work.

The natural reaction is to reach for your phone. You have AI, too. Why does theirs matter more than yours? Because consumer tools are general. They do not know your file, do not sit inside the institution’s systems, do not carry the memory of thousands of similar cases handled before yours. On the other side, AI is built into the systems lawyers already use, with trained staff to check its work. If your tool is wrong, you bear the cost. If theirs is, the institution absorbs it. The imbalance is not access to AI, but rather the infrastructure around it.

I wrote letters like the ones now arriving in your mailbox — insurance denials, eviction notices, collection demands — for institutions that paid me well to write them. These letters are not written to inform you. They are written to move you: toward a settlement you cannot evaluate, a deadline you cannot meet. When a person wrote them, they took hours. But the machine writes them in seconds and never tires. The denial is drafted instantly; the letter explaining why your reply was insufficient is drafted instantly too. The institution can now answer you far faster than you can answer it.

The risk has reached the courtroom. Last year, in Zhang v. Chen, the B.C. Supreme Court caught a lawyer citing two decisions invented by ChatGPT — cases that did not exist, generated by a machine, filed by a lawyer, read by a judge. An Ontario court caught a similar issue months later.

These trained, regulated, insured officers of the court got it wrong often enough that Ontario rewrote its rules in late 2024 to require lawyers to personally certify that every case they cite is real.

Now let’s go back to the woman at her kitchen table. Laid off after 23 years, with an insurance denial she cannot parse. No law school. No bar exam. No insurance. No colleague down the hall to check her work. Up against this system, she has no chance.

Imagine nothing changes. Courts and tribunals, already overloaded, slow further. More Canadians arrive without a lawyer; more give up before they begin. People begin to believe the justice system isn’t reachable at all.

Thankfully, the fixes are not mysterious or expensive.

Canadians have a right to be told, in plain language, when a machine helped produce the letter on their kitchen table.

The same tools the institutions use need to be funded into legal aid and the courts and put in the hands of the people who receive those letters.

And certification rules like Ontario’s should be matched, at the moment a document is filed, by automatic verification against the public databases the courts already trust.

A legal system earns its legitimacy from the promise that an ordinary person can survive contact with it. Not win. Not understand every step. Just survive, long enough for the truth of their situation to matter. If one side becomes industrial and the other stays human, that promise breaks.