Beyond Prompts: Why Legal AI Must Be Built Around Matters, Not Conversations
Conversations Are Stateless. Matters Are Not.
A chat interface resets. Every prompt begins a new context window. But a matter accumulates — documents, research threads, negotiation history, client instructions, prior drafts — over weeks or months. The fundamental mismatch between the conversational paradigm and the matter lifecycle is why early legal AI tools felt impressive in demos and brittle in practice.
What a Matter-Centric Architecture Looks Like
Irys organizes work at the matter level. When a lawyer opens a matter, the system already knows the governing jurisdiction, the relevant contracts and filings, the open research threads, and the last action taken. Prompts are contextualized automatically. There is no need to re-explain the client, the dispute, or the jurisdiction on every query. The AI has persistent context and uses it.
The Compound Effect of Persistent Context
Persistent context does more than reduce friction. It compounds. Early research on a matter informs later drafting. A precedent found in week two is automatically surfaced again when drafting the reply brief in week six. Anomalies in document sets that span multiple uploads can be flagged across the full corpus rather than within a single session. This is the difference between an AI tool and an AI colleague — the colleague remembers.
Implications for Knowledge Management
Law firms have invested heavily in knowledge management infrastructure — precedent databases, deal room archives, matter management systems. A matter-centric AI architecture can integrate with this infrastructure rather than sitting beside it. Irys is designed to ingest and index firm knowledge at the matter level, meaning institutional precedent becomes contextually available rather than requiring a separate search workflow.
What This Means for Junior Associates
For junior associates, the matter-centric model is transformative. Instead of spending the first two hours of every research session rebuilding context from prior memos and emails, the system surfaces the live state of the matter immediately. Research assignments become more targeted. Draft quality improves because the context is complete. The associate can spend more time on the judgment-intensive work that actually develops legal skill.
