I use AI. Here's how.

Summarizing medical records

My office requests the records, keeps a patient ledger of everything still owed to us, and follows up with anyone who hasn't sent it. That part is ours. What the tools do is read what comes back, index it and summarize it, so that by the time I sit down with a file I'm reading the pages that matter instead of hunting for them. On a serious injury case that's thousands of pages, and it used to be thousands of pages of billable review.

For a long time the only way to do that was a person, page by page, and the hours appeared on the invoice. There's another way now, and I'd rather not send you a bill for it.

Instead, I use a product designed specifically for law firms. It reads the entire record set and surfaces the things that decide these cases: the old injury to the same knee, the degenerative findings that predate the accident by a decade, the eight-month stretch with no treatment at all, the history given to one doctor that doesn't match the history given to the next. Then I can ask questions in plain language and get answers drawn from the records rather than from thin air.

Ten thousand pages of medical records, and the answer to almost any question I put to them in about twenty seconds. When did the plaintiff first complain of headaches? What was the total billed to Medicare? Is there a reference anywhere to a provider whose complete records I don't already have? Questions that once cost a day of paging through binders now cost less than a minute, and every answer still gets checked against the record before I rely on it.

In my firm, we don't outsource the collection of medical records. That's a deliberate choice, and here's why.

Record retrieval is often handed to a vendor, and I understand the appeal. But a vendor is paid to send a request and forward whatever comes back. No vendor has any reason to notice what didn't come back.

Every record that arrives here gets read for the names of other providers: another doctor, a lab, an imaging center, a therapist, an emergency room visit two years before the accident. If a record points to treatment we don't have, we chase it down. One set leads to the next until the trail runs out.

There's a reason I'm this stubborn about it. In almost every FELA and personal injury case I've tried, I've shown the jury far more of the plaintiff's medical records than the plaintiff's own lawyer did. In my experience, the complete medical file helps the defense more than it helps the plaintiff. In one admitted-liability FELA trial, the plaintiff had back surgery after a derailment and asked the jury to hold the railroad responsible for it. A single line in the surgical nursing notes told a different story: he had admitted to a nurse that he had needed that surgery for years and had simply been putting it off. The jury returned a defense verdict.

That's why, once a client agrees the full medical file is worth the effort, I don't consider the job finished until we have every record we know about.

Over the years I've been brought into many cases that another firm worked up first. When the record collection was outsourced, it shows, usually in the form of large holes in the medical history. So the tools carry the reading here, and the collecting stays with us.

Spotting issues in a file

Claude gives the record a second read and never gets tired doing it. Claude flags inconsistencies, gaps, dates that don't line up, and things worth a closer look. Claude raises the questions. Answering them, and deciding anything, stays with me.

First drafts

Letters, interrogatories, deposition outlines, deposition summaries, and the other documents in a case that follow a known shape. What comes back is a first draft and gets treated like one. Every word of it goes through me before it leaves the office.

Scheduling, invitations, and deadlines

Setting meetings and circulating calendar invitations. Docketing works the other way around. My staff and I calculate every deadline ourselves, from the rule and the order, and we're the ones who enter it and set the tickler behind it. Artificial intelligence runs second, as a check on us, re-reading the order and recomputing the date so Claude can flag a disagreement for one of us to resolve. Claude never sets a deadline and never puts anything on the docket.

Keeping the file moving

Claude also handles the paperwork that keeps a case moving. It puts documents on my letterhead and in the format the court requires. It prepares certificates of service and cover letters, and makes sure the right version of a document, PDF or Word, reaches the right person. It drafts my time entries the same day I do the work, instead of my rebuilding them from memory at the end of the month, which is all too common in law firms. And it finds things: a form I used two years ago, an exhibit buried in a stack of produced documents, the one email where someone agreed to something. You rarely see this work, but you pay for it, so the faster it gets done, the less it costs you.