■ CRITICAL RISK ■ Legal
The document-heavy core of paralegal work — review, research, drafting from templates — is being automated at speed, so yes, the role shrinks substantially. Paralegals who run cases, clients, and the AI tools themselves will still be billing; the ones who mainly summarized documents won't.
“AI reads contracts faster than you read emails.”
Our AI replacement risk score — how we score jobs
Paralegals are the leverage layer of law practice: conducting legal research, reviewing and summarizing documents in discovery, drafting routine filings and contracts from precedents, cite-checking briefs, managing case calendars and deadlines, organizing exhibits, and interfacing with clients on facts and paperwork. Firms bill their hours out at a healthy multiple of their salary, which is exactly why the economics of automating them are irresistible to clients demanding lower bills.
Legal tech got there early and LLMs supercharged it. E-discovery platforms with technology-assisted review have been culling document sets for over a decade — predictive coding was displacing armies of document reviewers before anyone said 'generative.' Now tools like Harvey, CoCounsel, and Lexis+ AI summarize depositions, draft discovery requests, run first-pass legal research with citations, and mark up contracts against playbooks in minutes. Cite-checking, privilege logging, deposition digesting, first drafts of routine motions — each was a classic paralegal assignment and each now has a credible AI version that a supervising attorney reviews. Firms aren't firing paralegals en masse; they're not replacing departures and not scaling headcount with caseload, which arrives at the same place slower.
The resilient work is procedural and human. Courts have local rules, cranky clerks, and filing quirks no model tracks reliably; hallucinated case citations have already gotten lawyers sanctioned, which makes verification — a paralegal skill — more valuable, not less. Witness coordination, client hand-holding, trial logistics, and knowing where everything is in a thousand-docket case remain stubbornly human. The paralegals who become their firm's legal-tech operators are gaining ground even as the traditional pyramid narrows beneath them. That narrowing is what 89 means.
Automatability: our editorial assessment of current and near-term AI capability
The pressure is current, not coming. E-discovery automation reshaped document review years ago, and generative tools moved into research, drafting, and summarization across large firms and legal departments in the past few years. Through the late 2020s expect flat-to-declining paralegal hiring even as caseloads grow, with the squeeze hitting document-review-heavy litigation support first and procedure-heavy trial and client work last.
It's a riskier one than it was. The research-and-document-review core that filled junior paralegal years is automating quickly, and firms are hiring fewer people to do it. The role survives in a leaner, more technical form — case management, court procedure, AI-tool operation, client coordination. Enter with eyes open, specialize early, and avoid stacking your career on pure document review.
First-pass document review in discovery went first, via predictive coding. Generative tools now handle deposition summaries, research memos with citations, contract review against standard playbooks, and drafts of routine motions and discovery requests. Everything still needs attorney review — hallucinated citations have produced real sanctions — but the drafting hours those tasks used to consume are collapsing.
No — but they'll hire differently. Firms still need people who navigate court procedure, wrangle clients and witnesses, manage cases, and operate the tech stack. What disappears is the leverage model of many junior paralegals doing document work under one attorney. Postings increasingly ask for e-discovery platform experience and AI-tool fluency alongside traditional skills. Fewer seats, higher bar per seat.
Three moves. Master the tools: e-discovery platforms, contract-review AI, research assistants — and become the person who validates their output. Own procedure: local rules, filing systems, and trial support are experience-based moats. And pick a specialty where facts matter as much as documents, like immigration or personal injury, since client-facing factual work automates slowest. Optionally, look at limited-license legal practitioner programs opening in several states.