Why this job keeps a person in the chair
Legal support work runs on deadlines set by other people. A filing window closes. A judge wants a specific format. A client calls back at 4:50 p.m. with a document that changes the pleading. Software can draft fast, but someone has to own the calendar, the court rules and the consequences of a late filing.
Two tasks show the split well. Preparing pleadings, correspondence and routine contracts from existing templates is pattern work, and language models handle patterns. Attending client interviews, hearings and depositions, taking notes and turning them into usable records is not pattern work. It needs presence, discretion and a judgment call about what matters.
There is a third thing that holds the role together: the file. Case files, exhibits and signed originals have to be organized, tracked and produced on request. Tools help with that, but the responsibility for an accurate, privileged, court-ready file sits with a named human at the firm.
What software drafts, what it assists with, and what stays with staff
Software already handles a slice of the day on its own. Drafting standard letters and template documents, and running database searches of statutes and decisions to pull candidate authorities, are the clearest examples. The task split above puts 13% of task time in that group. Those outputs still get checked before they leave the office.
A larger part of the work is assisted rather than handed over. Summarizing medical or employment records, cleaning up formatting, transcribing dictation, and sorting incoming correspondence all go faster with a tool in the loop. Assisted tasks account for 55% of task time. The pattern is familiar: the first draft gets cheaper, the review does not.
Then there is the part that stays with people: 32% of task time. That is the deposition and hearing attendance, the client contact, the court filings made under a specific clerk’s rules, and the calendaring that keeps a matter alive. On the headline measure, this job scores 66 out of 100 (higher is safer). Can AI do it? Our coverage figure, 38 on a 0-100 scale, is explained on the coverage method page.
What has actually been measured
Not much, for this job specifically. The evidence grade here is D, which means no study has tested an AI system against a working legal secretary on this job’s real tasks. Because of that, we publish no parity number. Is it better than a person? There is no honest answer yet, and our quality parity method explains why we leave it blank rather than guess.
What would settle it is specific: a timed test where qualified legal assistants and an AI system both prepare filings for real court rules, calendar a docket, and produce a case file, scored on accuracy, missed deadlines and rework. Benchmarks on legal research questions do not cover that, because research is only one strand of the job.
The outside numbers that do exist are about demand, not quality. BLS counts about 156,280 legal secretaries and administrative assistants in the US, with median pay of $55,570, and projects employment to fall roughly 5% between 2025 and 2035 (BLS, 2025). That is a shrinking entry door rather than a job disappearing. You can see how that compares across office roles in the secretaries and administrative assistants family.
The timeline, and what moves it
Most likely between 2036 and 2050 (8 in 10 of our scenarios). The replacement year method sets out exactly what that window measures and how it is built.
Two things could pull it earlier. Document assembly is already native to practice management systems, so adoption does not need a new vendor, just a setting turned on. And the cost gap is wide: the panel above puts annual software spend far below the cost of a salaried assistant, which is the kind of math partners act on when billable rates are under pressure.
Two things hold it back. Confidentiality and privilege rules make firms cautious about pushing client material through outside systems, and malpractice exposure makes a human signature on filings non-negotiable in practice. The physical side also resists: the robotics tier for this job is fixed automation, meaning scanners, copiers and mail equipment, not a machine that walks a courtesy copy to the clerk’s window.
Good to know: the roles thinning fastest are the ones built almost entirely on typing and formatting, not the ones holding the docket.
Three things to lean into
Pick the tasks that already sit in the human column. Own the docket and the court-rule calendar for your matters, including local and judge-specific requirements. Be the person who handles client contact when a case gets tense. Take the notes at hearings, interviews and depositions, and turn them into a record attorneys actually use.
Two skills pay off alongside that. The first is e-filing and court technology across the systems your firm touches, state and federal. The second is checking AI output for the things it gets wrong in legal work: invented citations, wrong jurisdiction, stale rules. Our guide to AI skills employers want covers the second in more detail.
If you are weighing a move, the closest work is next door. Paralegals and legal assistants trade formatting for substantive case work. Executive secretaries and executive administrative assistants lean on judgment and access rather than document volume. General secretaries and administrative assistants sit on a broader base of employers. You can put any two of them side by side on the job comparison tool, see where legal support sits among jobs most at risk, or read how every figure on this page is built on the methodology page. Firm-level context sits on the law firms sector page.