P2P

P E E R T O P E E R M A G A Z I N E · S P R I N G 2 0 2 6 49 breaks through this ceiling is the move from a single-skill chatbot to a unified, multi-capability assistant. In a single conversation, Athena can now handle drafting, document review, image and file analysis, web search, and internal resource lookups. This is not a cosmetic change. It fundamentally alters the user experience. Context is preserved across the conversation. A user can upload a contract, ask for a summary, request a comparison to a standard template stored in firm resources, and then ask who in the firm has handled similar agreements, all without leaving the chat. The assistant becomes a workflow, not just a tool. SCALING RESPONSIBLY: WHAT IT TAKES Building a multi-capability assistant is relatively straightforward. Scaling it into something attorneys trust and differently, had become a friction point. With intelligent routing, the assistant could determine the user's intent and dispatch to the appropriate underlying capability transparently, with no menu of options and no need for the user to understand the technical architecture. This orchestration is what allows Experience Connect to function reliably in practice. An attorney can upload a document, ask who has handled similar matters, and receive results grounded in governed firm experience data -- without needing to choose tools or adjust how the question is framed. When those answers are consistent and traceable to firm records, trust builds quickly, and expertise finding becomes part of the normal workflow rather than a separate step. The architectural leap that continue to use is not. That transition is where AI maturity is tested. Platforms tend to stall or advance based on three disciplines: routing precision, data quality, and feedback loops. As capabilities expand, the assistant must reliably interpret intent and route requests, often ambiguous ones, to the correct underlying tool, whether that is drafting, document analysis, or expertise finding through governed internal data such as Experience Connect. Getting this right is not a configuration exercise; it is an ongoing operational commitment. An example of this principle in action is the firm's introduction of a dedicated Westlaw Deep Research skill, accessed within the Athena interface. This skill is designed to route attorneys to the appropriate tool for AI-assisted legal research and delivers results separate from Athena's general chat. While Athena's general chat supports a wide range of tasks, it is not intended for legal research. The Westlaw Deep Research skill allows attorneys to reach authoritative, verifiable answers sourced directly from Westlaw through a familiar interface using streamlined workflows while reinforcing the firm's standards for responsible and accurate AI use. This kind of purpose- built routing, where the agentic assistant recognizes a legal research request and directs it to the right specialized tool, rather than attempting to answer from general knowledge, illustrates the

• Cover