Our Take
The real story is not architectural novelty but the shift from 'does AI work?' to 'how do we make AI safe to deploy?' — and iManage's answer is: lock down your documents first.
Why it matters
Law firms are moving past generative AI pilots into production use. Without a governed knowledge foundation, AI agents trained on firm documents become liability vectors. This matters now because the legal industry is operationalizing, not experimenting.
Do this week
Legal ops and knowledge managers: audit what documents and metadata your firm would expose to AI agents today, and flag governance gaps before vendors push integrations.
iManage Repositions as the Governance Layer for Legal AI
iManage, the document and knowledge management platform used across law firms, announced two technical moves in the weeks before this podcast episode. The first is a "context fabric," an architectural layer that organizes a firm's accumulated documents and activity logs into what the company calls a "living, governed foundation" for AI agents. The second is iManage MCP, an open-protocol connection that allows external AI systems to access iManage-stored content without custom point integrations.
In a conversation with LawNext host Bob Ambrogi, CEO and cofounder Neil Araujo frames the shift bluntly. Two years ago, law firms were feeling their way into generative AI. Now the question is operational: how do you actually deploy AI at scale while maintaining security, relevance, and trustworthiness of AI outputs.
Araujo's thesis rests on a single observation. Fluent AI responses are worthless if they cite the wrong contract or miss a conflict check buried in firm metadata. The foundation that makes AI safe is not the model itself but the context layer: which documents an AI can access, which it cannot, and what governance rules apply to each. Without that, AI becomes a fast way to generate plausible but unreliable answers.
Context Without Governance Scales Liability, Not Capability
Legal work is high-consequence. An AI hallucination in contract review, discovery response drafting, or legal research can expose a firm to malpractice or sanctions. The industry's prior AI experiments were largely constrained: a chatbot over a knowledge base, a contract analyzer on a closed set of documents.
The operational move iManage is betting on is different. As firms integrate AI agents into work streams, those agents need access to broader document sets. But broader access without governance is a risk multiplier. A firm that exposes its entire document repository to an AI model without controls over which documents are relevant, current, or legally privileged has not improved efficiency; it has widened the aperture for error.
Araujo's framing of "context fabric" as the foundation reflects this. The hard work is not the AI model. It is the metadata, the access rules, the versioning, the audit trails. That is where the liability sits, and that is what iManage is positioning itself to manage.
The MCP protocol announcement is a subsidiary move but tells a story. By offering an open connection standard rather than forcing firms into custom integrations, iManage is signaling that it expects multiple AI systems to touch firm documents. The governance problem is not "which AI vendor do we choose?" but "how do we govern access across multiple AI vendors at once?"
Know Your Exposure Before the Integration Wave
For legal operations teams and knowledge managers, the immediate action is inventory. You need to know what documents, metadata, and activity logs your firm would be exposing to AI agents if you enabled iManage MCP or similar integrations tomorrow. Which documents are current? Which are privileged? Which ones contain conflicts data? Which ones are subject to client confidentiality clauses that may restrict AI processing?
This is not a vendor selection problem. It is a governance readiness problem. Once you understand your exposure, you can enforce controls. Without that work, any AI integration at scale becomes a liability multiplication exercise rather than a productivity gain.