Contract Review & the AI-Native Model

TL:DR: General Legal's contract review runs on three layers: Sentinel, an internal AI that gives every contract a first pass in about ten seconds, attorneys who spend 15 to 30 minutes finishing the review instead of the several hours it used to take, and a hiring model built to attract the kind of high-leverage attorneys and engineers this workflow depends on. Independent research backs the scale of the speedup. Bloomberg Law's 2024 Contract Workflow Analysis found AI-assisted review cut standard contract review time from 92 minutes to 22 minutes, a 76% reduction, and a GC AI survey of active users reported saving an average of 14 hours per week. The model isn't AI replacing lawyers. It's AI doing the parts it's good at, issue spotting and drafting, so attorneys can spend their time on the parts it isn't: interpreting client needs, judging what's market, and setting deal strategy.

The Technology: Sentinel and a Slack-Native Workflow

Most of our clients work directly with us over Slack. Each client's private Slack channel is monitored by bots that detect new client requests, trigger our contract AI workflows, and schedule attorney time for review.

How Sentinel Is Built

Our contract AI (which we call Sentinel) is built on a few simple ideas:

1. State of the art LLMs are good at issue detection, redlining, and annotation
2. The best LLM agents are bad at interpreting client needs, leverage, and defining strategy for deals
3. LLM world models do not know what's market and LLMs are both overconfident in their market knowledge and over-aggressive in their mark-ups
4. The best human attorneys are really good at 2 and 3 and often not particularly great at 1.

That gap between what AI does well and what attorneys do well isn't just General Legal's internal read. A LexCheck 2024 benchmark testing clause identification across 12 standard categories, including indemnification, limitation of liability, and IP ownership, found AI tools reached 94-97% accuracy, while experienced lawyers reviewing the same contracts averaged around 80%, with accuracy dropping further under time pressure. That's exactly idea 1 and idea 4 above, measured independently.

The Internal Review Workflow

We turned these ideas into General Legal's core internal contract review workflow:

1. Sentinel AI always takes the first pass over a contract, guided optionally by an attorney defined strategy
2. Sentinel creates redlines, annotations, and issues lists. These lists are overly inclusive, sometimes misinterpret market, and don't fully align to the client strategy
3. After Sentinel finishes its review (in about ten seconds), our attorneys take over reviewing the full contract as well as the AI's markup.

Sentinel AI reviews the contract guided by an attorney defined strategy

Sentinel generates redlines, annotations, and issues lists in about 10 seconds

The Humans: Where Attorneys Add the Value AI Can't

Experienced attorneys can typically do a post-Sentinel review in about 15-30 minutes, a compression of several hours of formerly billable work. The AI related human acceleration comes from three sources:

1. having all issues clearly marked in the document with concise explanations
2. having many redlines and comments pre-built that the attorney can edit
3. having full client context (including prior contracts) in AI memory and easily accessible to the attorney through our tooling

Human acceleration also comes from several non-AI factors built into our workflow:

1. Contract review happens in only one place, Microsoft Word. This keeps the attorney fully focused on the actual contract document that gets turned back to the client. We built a specialized Word add-in to add additional AI capabilities to the review without pulling the attorney into another tool
2. Great UX is the standard for any tool an attorney uses
3. Automation around every handoff point. Bots and agents manage our scheduling, and notification, eliminating hours of lost productivity in the email driven workflows of traditional law firm contract review

For context on how much time is actually at stake, broader industry data points the same direction: a 2026 Forrester Consulting Total Economic Impact study of Thomson Reuters CoCounsel found customers achieved up to a 33% reduction in time spent on document review, research, and drafting.

The Talent: Hiring for a High-Leverage Model

The best traditional law firms are phenomenal managers of talent. Talent is core to the identity of law firms, built into their apprenticeship models, and a point of fierce competition. We think the best AI-native law firms will be the ones that can also hire the best attorneys and AI engineers. Good engineers and good attorneys already deliver 10x+ leverage multiples over bad ones. In the AI-native model, the leverage of these already high-leverage people can be massively compounded. Well-run AI-native companies should trend toward at least 100x employee leverage.

Two Lessons from Traditional Law Firms

How does a new AI-native law firm win the kind of talent that can deliver those multiples? We're drawing a couple lessons from traditional law firms:

1. Hire people to work on something that matters. Law matters and good lawyers became lawyers to work on important things.
2. Giving real ownership stakes to top talent yields exceptional outcomes.