For Nadia Louis-Hermez at NEXT Insurance, part of the answer is policy maintenance. During the “From Prompt to Process” panel, she described Ask Vinny, a legal front door built with Checkbox and available through Slack. It draws on supplied knowledge to answer routine questions, directs unresolved requests into tickets and supports escalation to people when review or approval is required.

Employees gain access to guidance without someone in legal fielding every initial inquiry. The department assumes a continuing obligation to keep that guidance current. As Louis-Hermez explained, a policy change has to reach the system’s knowledge source; otherwise, employees can receive an answer grounded in a version the company has superseded.

At iFood, Heloisa Alves described the effort to make dispersed knowledge usable across the department. Requests had arrived through direct messages and channels, reaching different lawyers and sometimes producing different answers. Centralized intake and a shared knowledge base gave AI agents access to policies, guidance and recurring answers contributed by the practice areas.

An audience member questioned whether that would ensure consistency. Alves acknowledged the limits: cases and lawyers differ, and AI cannot guarantee identical responses. The objective was a common foundation that helped routine questions move while preserving lawyers’ attention for work requiring judgment.

That distinction gives the knowledge itself greater importance. Advice carries assumptions about facts, circumstances and acceptable risk. Reusing an answer requires enough context to distinguish a position the company routinely takes from an exception it approved once. The answer may travel more easily than its qualifications.

AI extraction helped fill those gaps, with people monitoring the results. Before colleagues could confidently ask what the contracts said about a particular issue, the department had to improve the information on which the response would depend.

These are manifestations of a broader problem. Gartner’s July 2024 survey of 1,203 data management leaders found that 63 percent of organizations either lacked the right data management practices for AI or were unsure whether they had them. Publishing the findings in February 2025, Gartner emphasized the need for continuing investment in metadata, monitoring and governance. 

The panel showed what that investment entails inside legal. Some of it repairs deficiencies that predate AI. Some enables questions the department previously struggled to answer. Its return can include both less repetitive work and greater capacity to examine what the company already knows.

The speakers did not provide a full accounting of those costs and gains. Their examples nevertheless make a narrow measure of productivity harder to defend. Counting completed requests captures the answer delivered. It may miss the work that makes the next hundred answers possible.

For general counsel, the investment decision extends beyond the technology. The people maintaining legal knowledge contribute to every task that draws on it. Their work belongs in the cost of AI—and in the credit for its return.