Why AI Might Actually Create More Work for Lawyers

Lowenstein Sandler has roughly 400 lawyers concentrated in the New York area serving private equity, venture, hedge funds, technology companies, and life sciences clients. Despite a 15-year movement toward alternative fee arrangements, project-based and value-based pricing has not taken hold in corporate law, which Gary Wiggins attributes partly to mutual distrust between lawyers and clients. Law remains one of the last professional services segments still organized around billable hours, even though no client has ever asked to purchase billable hours as such.

Wiggins frames AI as operating along two distinct dimensions that separate it from prior legal technology. Earlier tools like blacklining software addressed only efficiency by compressing a manual task to roughly 90 seconds. AI functions instead as a thought partner that improves work product from the outset. A partner specializing in international tax uses AI to test structuring theories, then assigns an associate to verify the output through traditional research, with the iterative process producing structures neither the human nor the AI would have reached independently. Wiggins acknowledges difficulty in objectively measuring whether this co-pilot dynamic genuinely adds insight or creates an illusion of new thinking.

Concrete results are already visible in two practice areas. Patent lawyers using an AI drafting tool for just over six months report it produces broader applications by drawing on knowledge across all engineering and scientific disciplines, and clients unprompted told the firm they noticed improved patent quality over that same period. On the transactional side, AI reduced the cost of one due diligence project by 70 percent by replacing the first layer of document review across thousands of trust agreements with a 100-column spreadsheet that lawyers then quality-checked. Clients still pay law firms to certify that output because they lack the staffing and at-the-market knowledge to do it themselves, particularly for transactions that are one-time events for them.

The Jevons Paradox is the central economic argument in the episode. A document review job quoted at a hypothetical ten million dollars generated zero revenue when declined, but the same job priced at three million dollars after AI efficiency gains generated three million. Reducing discovery costs from roughly two million dollars to two hundred thousand dollars changes the economic calculus for filing lawsuits entirely. One client reported a quadrupling in patent invention requests from its own engineering team after adopting AI internally. The prediction is that lower per-unit legal costs will generate enough additional legal activity to increase total lawyer workload, preserving the profession's economic position overall. Larger clients with dedicated legal teams are expected to keep more routine work in-house as AI automates grunt work, following the pattern of banks bringing derivatives documentation in-house, while outside firms retain value where market breadth and independent perspective are harder to replicate internally.

The firm has not changed its hiring patterns and is filling a full slate of first-year associates and summer 2027 positions, but it is changing the skills it looks for and how it trains new lawyers. Junior lawyers now bring laptops with AI tools open during first-year curriculum exercises, including prompting AI with deal provisions and evaluating seller-favorable versus buyer-favorable outputs. A genuine concern exists that removing tedious tasks like blacklining eliminates repetitive work that built deep skills the way musicians build technique through scales, and the profession has not resolved how to develop lawyers without that foundational grunt work.

Harvey and Ligora are described as the two dominant AI platforms in law firms, with nearly all lawyers at Lowenstein Sandler using Harvey regularly. Harvey differentiates from consumer tools like Claude or ChatGPT by adding a security layer that prevents client data from reaching the internet and providing retrieval-augmented generation trained on legal content. Using consumer-grade models for legal work can cause lawyers to lose attorney-client privilege, making enterprise platforms a compliance necessity. The regulatory environment has shifted sharply: two years ago malpractice carriers flagged AI use as a risk concern, and now they treat AI competence as an assumption. Clients who 18 months ago told firms not to use AI are now demanding it to reduce costs, though lawyers must still verify all output given documented cases of briefs citing hallucinated cases.

Average hourly rates at the largest US law firms rose 10.1 percent in 2025 against roughly 3 percent CPI inflation, a gap described as historically unprecedented. The explanation offered is that AI makes each billable hour more productive and therefore more valuable, so rates rise even as total hours per matter fall and clients pay less overall per solution. Kirkland and Ellis announced 500 million dollars over five years on AI-related technology, though training a foundation model from scratch costs approximately 1.5 billion dollars, meaning even that commitment falls well short of building proprietary models. Lowenstein Sandler has decided not to train its own models and will instead customize on top of existing ones, including building a firm-specific M&A playbook. A pricing risk flagged for the industry is that enterprise AI tools currently priced on all-you-can-eat models are shifting toward token-based pricing, and current rates may be subsidized by outside investors in a pattern that could mirror the eventual repricing seen with Uber and food delivery platforms.