AI in the courts stopped being a question of principle this week and turned into a list of three tools. Sarah Sackman, the justice minister, told a City of London Law Society conference that she does not want a two-tier system. In that system the largest firms work at the front of AI use while the courts sit, in her words, stuck in the analogue age. The day before, writing under her own name, she set out what the government is building for the court estate, being transcription, listing and case summarisation. Read those three words as a work plan rather than as a speech.
What the minister has promised
The two-tier warning is the line that will be reported and the smaller part of the story. Transcription produced by machine changes what the record of a hearing is and how quickly you get hold of it. Listing touched by software changes how a hearing date is chosen and how much notice you receive. Case summarisation puts a machine-written account of a dispute into the system that handles your client's file. None of it is running yet and the minister gave no date for any of it, so what you have is a direction rather than a deadline. It comes from a department that has already put legal services first in line for the AI Growth Lab sandbox, where applications close on 27 September.
A faster firm and a slower court
A practice that drafts in half the time still waits the same number of weeks for a hearing. That mismatch does not stay with the court. It lands on you, because a client who paid for speed reads every later delay as yours. The answer is not to slow down to match the list. It is to say plainly, when you take instructions, which part of the timetable your firm controls and which part it does not, and to put that in the client care letter rather than into a difficult telephone call nine months later.
The second effect falls on the record itself. Where a transcript arrives from a machine, somebody has to read it against the note taken on the day, and that somebody has to be the person who was in the room. Where a summary of the case is written by software, your own chronology is the only thing you will have to correct it with. A firm that keeps a running chronology on every contested file will find this cheap. A firm that rebuilds the story from the correspondence each time it needs one will find it expensive, and will find it out late.
What to settle this month
Three decisions sit better made now than in a rush. Name the person who checks a court-produced transcript against your note, and give them a deadline measured in days. Write down what your firm does when a court document misstates the case, including who signs the letter and how quickly it goes out. Then look at the AI Growth Lab before it closes on 27 September, because the sandbox is open to law firms and conveyancing businesses as well as to software companies. A practice with a real problem to test has a stronger application in it than most.
Watch what the department funds rather than what the minister says. A transcription contract with a start date tells you more about next year than any conference platform will.
Sarah Sackman set out the government's case in her own words on 15 September, and her article is open to any reader at City AM.
If you want the gap between your firm's speed and the court's timetable read by someone who has sat on both sides of it, start with a conversation.
