When Anthropic announced its Claude Cowork plugins this week, legal technology stocks dropped. Thomson Reuters, LegalZoom, Wolters Kluwer—investors saw the writing on the wall. If an AI agent can draft contracts, review compliance documents, and build workflows without per-seat licensing, the traditional legal software business model starts to look fragile.1
The anxiety is understandable. Cowork represents a genuine shift in how legal work gets done.
But here’s the thing: we’re not worried. And it’s not because we think AI won’t transform legal work. It’s because Statvis solves a fundamentally different problem.
Generation vs. retrieval
Cowork and similar tools are generative. They produce text based on patterns learned during training. Need a contract clause? The model generates one that looks like the thousands of clauses it’s seen before. Need a compliance checklist? It synthesizes one from its training data.
This works brilliantly for tasks where “plausible and professionally formatted” is the goal. Contract drafting, legal research summaries, workflow automation—these are pattern-matching problems, and large language models excel at pattern-matching.
Environmental liability is different.
When you’re investigating a contaminated site, you don’t need an AI to generate a plausible site history. You need to know what actually happened, when it happened, and which document on which page proves it. The question isn’t “what might a site history look like?” It’s “what does this site’s history look like, based on these 14,000 documents?”
That’s a retrieval problem. And retrieval requires infrastructure that generative models don’t have.
Why citations aren’t optional
In environmental litigation, many of the claims will be challenged. When opposing counsel asks where you got the date of a 1997 release, “the AI told me” isn’t an answer. You need to point to page 47 of the Phase II ESA, paragraph three, where the field technician documented the discovery.
Statvis extracts events from your documents and cites every source. Each timeline entry links to the specific page and passage it came from. Click through, verify, defend your analysis. That traceability isn’t a feature—it’s the entire point.
Generative tools can produce citations, but they’re generating those too. The model predicts what a citation should look like based on context. Sometimes it’s right. Sometimes it invents case law that doesn’t exist. For contract drafting, you can catch and fix those errors. For establishing liability across decades of site history, that’s a reputational risk environmental experts can’t afford to take.
Different problems, different tools
Cowork will likely transform how law firms draft documents and manage workflows. That’s genuinely valuable work, and Anthropic has built impressive technology.
But environmental consultants and environmental lawyers investigating contaminated sites need something else: a system that processes their actual documents, extracts verifiable facts, and maintains the evidentiary chain that regulatory and legal scrutiny demands.
That’s what we built. And that’s why we’re watching the Cowork news with interest rather than anxiety.
Footnotes
-
Financial Times. “Software companies reacted swiftly to Anthropic’s new Claude Cowork plugins.” January 2025. ↩