Casepoint Adds AI Agents for Document Relevance and Issue Coding

Casepoint has deployed two specialized AI agents in its eDiscovery application, giving early-access customers automated support for document relevance and issue coding as of Sept. 23, 2026. The release expands Casepoint IQ, the company’s recently introduced intelligence layer, from a platform concept into customer-facing workflow tools. Both agents remain subject to human validation, recorded decisions and customer access controls.

The move reflects a strategic choice increasingly visible in legal technology: build narrowly scoped agents around repeatable work instead of asking one general-purpose system to manage an entire matter. Casepoint IQ brings together agentic AI, assistive AI, machine learning, predictive coding and MCP enablement across the company’s unified platform. That positioning matters because enterprise legal and government buyers must evaluate not only output quality, but also where data resides, how permissions apply and whether reviewers can reconstruct an AI-assisted decision.

Casepoint’s Relevance Determination Agent assesses whether documents meet a matter’s relevance or responsiveness criteria, explains its conclusion and attaches a probability estimate. Its companion Issue Coding Agent tests documents against legal, factual or investigative categories established for a case. A multi-model process cross-checks results before review. Customers first validate each agent against a sample, then can examine recall, precision and projected time and cost effects before expanding the run. The system preserves both human and machine decisions in an audit trail, while an optional quality check calculates accuracy for each issue code. Casepoint CEO Paul Colangelo said the company is developing additional agents for legal, investigative and information-governance work. Founder and Chief Technology Officer Vishal Rajpara said the design is intended to expose reasoning and stop for human approval at key stages.

Industry Implications

Casepoint is competing on operational control as much as automation. The company’s agents run inside its existing security boundary and follow customer permissions, governance rules and access controls, an approach aimed at regulated organizations that may resist detached AI applications. For legal operations teams, the significance is the workflow model: relevance analysis could precede issue identification and later data-handling tasks without forcing users across separate systems. Competitors will face pressure to show comparable evidence of performance, auditability and reviewer control rather than relying on broad claims about generative AI. Enterprise buyers, meanwhile, can assess agents as modular components, asking whether each task produces measurable gains and fits existing eDiscovery, investigations, legal hold, FOIA and compliance processes.

DreamLegal Perspective

Legal technology vendors should treat specialized agents as governed workflow products, not simply new interfaces for language models. The critical differentiators will be task-level accuracy, transparent testing, permission-aware deployment and the ability to connect multiple agents without weakening accountability. Legal operations leaders should demand baseline samples, recall and precision results, cost assumptions and a complete decision history before scaling automated review. Casepoint’s launch also gives the market a useful test: whether a portfolio of constrained agents can deliver more dependable enterprise value than a single broadly capable assistant.