AI in Legal: How Law Firms and Corporate Teams Use AI

60 of the AmLaw 100 firms now use Harvey AI. That is 60% of the largest law firms in the United States, deployed on a platform that handles legal research, M&A due diligence, and contract analysis — in a profession that was still largely using keyword search tools three years ago.

AI legal automation is happening on both sides of the legal relationship simultaneously. Law firms are using AI to handle more work per attorney. Corporate legal departments are using it to bring work in-house that previously went to outside counsel. Both moves are reshaping the economics of legal services faster than most law school curricula are acknowledging.

Here is what AI legal automation actually looks like across the five use cases that are delivering measurable value right now.

Why Legal AI Adoption Is Running at Two Speeds

In-house teams move faster. A corporate GC with eight lawyers managing 10,000 contracts per year has a math problem that AI solves directly. The outcome is clear: handle more work in-house, cut outside counsel spend, clear the backlog. No one loses a job — the team just covers more ground.

Law firms move slower, and there is a structural reason. The billable hour model creates direct tension with efficiency gains. If an associate used to spend 6 hours researching a brief and now does it in 90 minutes with AI, the client saves money — but the firm's revenue on that task drops 75%. Law firms that benefit from AI need to rethink pricing at the same time they deploy the tools, and most are not ready for that conversation.

The firms breaking through this tension are not doing the same work cheaper. They are using AI to take on significantly more clients and more complex matters per attorney. AI compresses commodity work so senior attorneys can sell more judgment work.

Goldman Sachs estimates 44% of legal work tasks are automatable with current AI. The question is not whether AI will change legal economics — it is whether firms are positioning to capture the efficiency upside or just absorbing the margin compression.

The Five Legal AI Use Cases Delivering Real ROI

1. Legal Research

This is where legal AI deployment is most mature. Tools like Thomson Reuters CoCounsel, Harvey Knowledge, and Westlaw AI allow attorneys to ask plain-language questions across case law, statutes, and regulatory guidance — and get back cited answers in under a minute rather than the 2-4 hour sessions traditional legal research required.

One litigation boutique cut brief research time from 18 hours per motion to 4 hours after deploying AI research tools — without any reduction in citation accuracy. That is not a marginal gain; it changes how the firm can price litigation work and how many matters a single team can carry.

The critical workflow requirement: every AI-generated citation must be verified against the actual source before it appears in any filing. Courts have already sanctioned attorneys for submitting AI-hallucinated case citations. AI research tools are fast and generally accurate, but they are not infallible — and the professional responsibility for accuracy stays with the attorney.

2. M&A Due Diligence

A mid-market M&A transaction generates 5,000 to 25,000 documents in the data room. Traditional due diligence means associates reviewing that volume under time pressure, checking 30-50 standard issue categories across the target's contracts, IP portfolio, employment agreements, and litigation history.

AI due diligence tools — Luminance and Harvey Vault are the two platforms with the deepest penetration among large firms — extract and flag issues across all issue categories simultaneously, in hours rather than weeks. A PE firm running a $200M acquisition can now complete contract-level due diligence in 3-4 days instead of 3-4 weeks.

The downstream business impact matters as much as the time savings: faster close timelines, lower transaction costs, and the ability to run more deals simultaneously without scaling associate headcount proportionally. For law firms competing on M&A mandates, the ability to turn around due diligence reports faster than the competition is a material differentiator.

3. E-Discovery

E-discovery is a $15 billion market built on manual document review — attorneys and paralegals reading millions of emails and records to determine relevance and privilege in litigation. It is slow, expensive, and introduces inconsistency as reviewer fatigue sets in.

Technology-assisted review (predictive coding) has been court-approved since 2012, but the current generation of AI-powered e-discovery tools represents a step change in capability. Modern systems process millions of documents, identify privileged material, cluster by topic, and surface the key 2-5% of documents that actually matter — in hours rather than months.

The cost reduction is real and significant. Document review that runs $1-3 per document under manual review can be processed at pennies per document with AI. For a case with 2 million documents, that difference represents $2-5 million in review costs. For litigation departments and the clients funding them, this is not incremental improvement — it changes whether certain cases are economically viable to litigate.

4. Contract Analysis

For a full breakdown of how AI contract review works — including clause extraction, risk scoring, and NLP pipeline — see our AI Contract Review deep-dive. The short version: AI reviews standard contracts at 94% accuracy versus 85% for experienced attorneys under time pressure, according to LawGeex benchmarks, while cutting review time by 70-85%.

Beyond review, AI contract drafting is gaining traction in enterprise legal departments. Tools like Spellbook integrate into Word and generate first drafts of standard agreements — NDAs, service agreements, employment contracts — based on party parameters. The realistic time reduction on routine first drafts is 60-80%.

The limit holds consistently: AI drafting handles standard terms well and novel deal structures poorly. Complex commercial agreements still require substantial attorney judgment. But the 80% of contracts that follow predictable patterns are faster and more complete with AI than without it.

5. Compliance Monitoring

Regulatory monitoring is one of the least visible but highest-stakes applications of legal AI. For companies operating across multiple jurisdictions — financial services, healthcare, pharmaceuticals, financial technology — the volume of regulatory change is impossible to track manually without a dedicated team.

AI compliance monitoring systems ingest regulatory feeds across jurisdictions, identify relevant changes, map them to affected business processes, and surface action items for legal and compliance teams. The Brinks legal team, for example, used Thomson Reuters CoCounsel to streamline global regulatory monitoring, cutting outside counsel spend while strengthening compliance coverage.

The ROI case is asymmetric in a way that makes the math obvious: a single material compliance failure — a GDPR violation, an SEC disclosure miss, a banking regulation breach — can cost more than the entire compliance team's annual budget. AI monitoring that catches a relevant regulatory update before it becomes a violation is not an efficiency play. It is risk management with a very clear cost-of-failure benchmark.

What the Teams Winning With Legal AI Are Doing Differently

They start with high-volume, low-complexity work. Contract review, NDA analysis, research memos for recurring question types. These build confidence in AI output quality before moving to high-stakes judgment work.

They build verification into the workflow. Not as optional review, but as the standard process. Every AI citation checked against source. Every AI-flagged clause reviewed in context. The goal is not to remove attorneys — it is to remove the portion of legal work that does not require attorney judgment.

They renegotiate pricing before AI pays off fully. For law firms, this is the hardest change. Firms moving fastest toward AI-driven economics are shifting toward flat-fee pricing on AI-enabled work — capturing some efficiency gain as margin while passing some to clients. Defending hourly billing rates for research and document review is increasingly difficult when AI compresses those hours by 70%.

FAQ

Which AI tools are leading law firms actually using?

The most widely deployed legal AI platforms in AmLaw 100 firms are Harvey (60+ of the top 100 US firms), Thomson Reuters CoCounsel (integrated with Westlaw and Practical Law), and Luminance (strong in due diligence and contract management). Most large firms run two or three tools simultaneously — platforms tend to specialize by use case rather than covering everything equally well. Firms also deploy Microsoft Copilot through existing M365 licenses for drafting and document work.

Does using AI create liability risk for attorneys?

Yes, and courts are enforcing this. Several attorneys have already been sanctioned and fined for submitting AI-hallucinated case citations in court filings. The professional responsibility rules have not changed: attorneys remain responsible for the accuracy of their work regardless of the tool that produced the draft. Any legal AI workflow needs mandatory verification steps built in — not as optional quality control, but as required practice before anything AI-generated leaves the firm.

Can in-house legal teams use AI to reduce outside counsel spend?

Consistently yes. In-house teams using AI for contract review, research, and compliance monitoring typically report 20-40% reductions in outside counsel spend on work that can be handled internally. The key is mapping which matters genuinely benefit from external specialization — complex litigation, novel transactions, high-stakes regulatory investigations — versus which are high-volume routine work that an AI-equipped in-house team can absorb without sacrificing quality.