Our Take
Centari is solving a real attorney workflow problem—amendments that shift deal meaning—but the win lives at the data layer, not the UI: firms get structured, trustworthy deal data they can port to any AI system, not lock-in.
Why it matters
Transactional lawyers spend hours reconstructing relationships between agreements, amendments, and schedules by hand. Centari's approach—auto-generating those maps and tracking how amendments change operative terms—cuts that toil, but only if firms commit to using Centari's data infrastructure as their source of truth.
Do this week
Deal teams: trial Amendment Awareness on a recent close with 3+ amendments (LPA, purchase agreement, or credit facility) to measure hours saved on data reconciliation before committing to platform adoption.
Centari Ships Amendment Tracking and Relationship Mapping
Centari, an AI platform for managing complex transactions, launched two features today: Amendment Awareness and Deal Maps. Both are available immediately to existing customers including Ropes & Gray, Willkie, Fried Frank, and Wilson Sonsini.
Amendment Awareness automatically updates a deal's underlying data when amendments are uploaded. Rather than tracking textual changes (redlining), the system reads how an amendment alters the meaning of the entire deal over time. The feature targets documents like amended and restated limited partnership agreements and amended purchase agreements, where rights shift between versions. Centari says the capability is most useful for credit facilities, portfolio company governance documents, and fund structures that evolve across multiple amendments.
Deal Maps generates a visual graph showing how documents in a transaction relate to one another, automatically depicting which agreements incorporate, modify, satisfy, replace, or supersede others. No manual classification or chart-building required. In a demonstration, the feature displayed a six-agreement M&A closing set where an amended and restated purchase agreement supersedes the original, with an equity commitment letter, reps and warranties insurance, and escrow agreement shown as satisfying or adding detail.
The Data Layer, Not the Application Layer
Kevin Walker, Centari founder and CEO (former Paul Hastings M&A attorney), positioned the company's focus differently from competitors like Harvey and Legora. Centari is building at the data layer—helping firms create high-quality, structured data assets from their own deal documents that firms can then use in whatever AI system they prefer.
Walker said the company is working toward a definitive ontology of a firm's knowledge and deal history, running from the practice group to the individual deal, clause, and data point. This framing sidesteps vendor lock-in concerns: you own the data. But it also means the value proposition depends on adoption depth. A firm that uses Centari only to generate Deal Maps for visual reference gets less return than one that treats Amendment Awareness output as the canonical source of deal terms.
Both features grew out of the founders' experience as practicing attorneys. Centari has expanded its team since late 2025, hiring attorneys from large firms and investing in applied legal research.
What to Test and What to Watch
Amendment Awareness solves a real problem: LLMs read documents as static text, while transactional lawyers track how agreements modify one another over time. The feature's ability to update deal data points automatically (not just flag text changes) could save significant hours on closing checklists and portfolio updates.
Deal Maps automates work lawyers currently do by hand or via closing checklists. The automatic generation without manual classification is the efficiency play, but only if the output is accurate and complete enough to reduce rework.
The catch: both features are only as useful as the data they produce. If Amendment Awareness misses a subtle modification to operative terms, or if Deal Maps miscategories relationships, firms inherit bad data. Centari's bet is that its "Deal Reasoning Engine" (described as patent-pending) is capable enough that firms will trust it. That claim remains unaudited by independent practitioners.