News You Can Use

Sequoia partner Julien Bek estimates $60 billion in externally handled legal work can be absorbed by AI autopilots: $36 billion in paralegal and LPO work, and $20-25 billion in legal transactional work such as contract drafting, NDAs, and regulatory filings. The framework draws a sharp distinction between copilots and autopilots. Copilots are tools sold to professionals - Harvey's current model, where the lawyer still does the work but faster. Autopilots are platforms that sell outcomes directly to end customers - Crosby drafting NDAs for companies, Lawhive providing legal services without a traditional firm in the middle. The core logic: if work is already outsourced, the buyer is already purchasing an outcome, not a relationship. And "today's judgment will become tomorrow's intelligence" as these systems accumulate domain-specific data, meaning the addressable market expands over time, not contracts.

This is the most important strategic frame published this fortnight, and it deserves serious attention from anyone running an innovation function in a law firm. The distinction between copilots and autopilots is not academic. Copilots reinforce the current model - lawyers do the work, just faster. Autopilots bypass it. If the $36 billion paralegal and LPO estimate is even half right, that is work that currently flows through law firms and ALSPs being rerouted to platforms. Crosby's $60M Series B this same fortnight, led by Sequoia and Lux, is the thesis being put into practice. The uncomfortable question for law firms is not "should we adopt AI tools?" - most already are. It is "what happens when the work our clients currently send us gets absorbed by platforms that sell the outcome directly?" The answer requires thinking about value differently: not hours delivered, but judgment, risk management, and the things a platform cannot replicate. The firms that can articulate that distinction clearly - to their clients and to themselves - are the ones with a strategy. The ones that cannot are relying on inertia.

Anthropic interviewed 81,000 Claude users across 159 countries and 70 languages over a single week in December - with Claude itself acting as the structured research instrument. The top hope across all respondents: professional excellence (18.8%). Not automation, not cost savings - being better at what they already do. The top fear: not job loss, as most commentary assumes. It is unreliability (26.7%) - people are worried AI will give them confident-sounding wrong answers. Two-thirds of respondents globally were positive about AI, and no country fell below 60%. But the regional variation is striking: Sub-Saharan Africa and South/Southeast Asia are markedly more optimistic than Western Europe and North America. The methodology alone deserves attention. Conducting 81,000 structured qualitative interviews in a week - something no human research team could attempt - demonstrates a use case for AI that goes well beyond drafting and summarisation.

Two findings matter most for anyone running AI adoption in a professional services firm. First, the fear data. Unreliability outranking job loss by a significant margin matches what we see internally and with clients. The resistance to AI is not existential anxiety about replacement. It is practical anxiety about trusting output that looks authoritative but might be wrong. That is a solvable problem - through better verification workflows, clearer confidence signalling, and honest training about what AI gets wrong - but only if you name it correctly. Framing adoption challenges as "change resistance" misses the point. People are not resisting change. They are resisting risk. Second, the professional excellence finding. When 81,000 people say their top hope for AI is being better at their jobs, that should shape how innovation teams pitch AI internally. The message that lands is not "this will make you more efficient" or "this will save the firm money." It is "this will help you do better work." That is a different conversation entirely, and it is the one most adoption programmes are not having.

Factor surveyed 204 in-house and law firm professionals for its GenAI in Legal Benchmarking Report 2026. AI access is now near-universal at 82.7%, up from 61.2% in 2025. But trust has not kept pace - only 22.1% report high trust in AI outputs. The headline finding: high-trust teams are three times more likely to report positive ROI from their AI investments. Over a quarter of organisations have spent between $100K and $500K on domain-specific legal AI. Law firms lead in-house teams on trust, confidence, and ROI - partly driven by client RFP pressure forcing firms to demonstrate competence rather than just access. Only 12.1% rate themselves as "leading the way." The access gap has closed. The trust gap is the new bottleneck.

This report should be required reading for anyone preparing an AI business case or reporting to a governance committee. The 3x ROI multiplier for high-trust teams is the sharpest data point yet on where adoption programmes should focus their energy. Access is no longer the problem. Giving people a login does not generate returns. What generates returns is the work that happens after the login: training that builds genuine confidence, verification processes that give people permission to rely on AI output, governance frameworks that make the boundaries clear, and measurement that shows results honestly. The RFP effect is worth noting for UK firms specifically. Clients are asking about AI capability in pitches and panel reviews. That external pressure is doing more to drive trust-building than most internal programmes - because it ties AI competence directly to revenue. If your adoption programme is still focused on rolling out tools and measuring logins, this report is the evidence that the strategy needs to shift. The firms reporting positive ROI are not the ones with the most AI tools. They are the ones whose people actually trust the tools they have.