Show: Beyond the Pilot · Publisher: VentureBeat · Host: Matt Marshall, Sam Witteveen
Episode URL: https://traffic.megaphone.fm/UTEAU4952214811.mp3
Publish date: 2026-02-18
Duration: NAs
Default source credibility: HIGH — Named F500 practitioners on-record with production metrics. VentureBeat editorial vetting. Treat vendor-sponsored segments as MEDIUM.
- LexisNexis, a legal information provider, faces challenges from AI agents and evolves its AI strategy to maintain competitive advantage.
- Min Chen, LexisNexis’ Chief AI Officer, discusses the transition from standard RAG to graph RAG to ensure authoritative legal responses.
- The company’s AI validation process includes a mix of human and AI evaluations to ensure high-quality, comprehensive, and trustworthy legal answers.
Extracted quotes
| # | Credibility | Speaker | Org | Timestamp | Topic | Quote |
|---|---|---|---|---|---|---|
| 1 | HIGH | Min Chen (Chief AI Officer) | LexisNexis | 05:51 | 07-adoption-challenges | RAG is really the solution can try to mitigate the hallucination issue inherited by the larger language model. In 2023, when we initially launched our first flagship JNL product, Lexis Plus AI, as I just mentioned, we built a standard RAG framework. And that is powered by a hypersearch, combining both lexical and vector search. And this allows us to grant models responses in LexisNexis-trusted authoritative knowledge base, which can help to ensure their reliability from day one. |
| 2 | HIGH | Min Chen (Chief AI Officer) | LexisNexis | 11:13 | 07-adoption-challenges | We have this seven to eight submetrics and putting together that forms an overarching measure that we refer to as usefulness. So the reason we assess performance across all these dimensions is that an answer may be contactually relevant, yet to still be considered as unuseful. Also very important for another metrics is comprehensiveness. So this metrics is designed to evaluate whether our Gen-AI response fully address all aspects of users’ legal questions. |
| 3 | HIGH | Min Chen (Chief AI Officer) | LexisNexis | 22:12 | 07-adoption-challenges | We have seen 20% increase by using Planner Agent. Now, the other agent you mentioned about Reflection Agent, that’s for the use case of transactional document drafting. The agent can automatically dynamically criticize its initial draft and then incorporate feedback dynamically and then refine the draft in real time, and that is the reflection agent is significantly improving the comprehensiveness and the quality of the responses. |
Per-quote detail
1. Min Chen — LexisNexis (05:51)
RAG is really the solution can try to mitigate the hallucination issue inherited by the larger language model. In 2023, when we initially launched our first flagship JNL product, Lexis Plus AI, as I just mentioned, we built a standard RAG framework. And that is powered by a hypersearch, combining both lexical and vector search. And this allows us to grant models responses in LexisNexis-trusted authoritative knowledge base, which can help to ensure their reliability from day one.
- Credibility: HIGH — Named exec with specific claim and unscripted interview.
- Topic tag:
07-adoption-challenges
2. Min Chen — LexisNexis (11:13)
We have this seven to eight submetrics and putting together that forms an overarching measure that we refer to as usefulness. So the reason we assess performance across all these dimensions is that an answer may be contactually relevant, yet to still be considered as unuseful. Also very important for another metrics is comprehensiveness. So this metrics is designed to evaluate whether our Gen-AI response fully address all aspects of users’ legal questions.
- Credibility: HIGH — Named exec with specific claim and unscripted interview.
- Topic tag:
07-adoption-challenges
3. Min Chen — LexisNexis (22:12)
We have seen 20% increase by using Planner Agent. Now, the other agent you mentioned about Reflection Agent, that’s for the use case of transactional document drafting. The agent can automatically dynamically criticize its initial draft and then incorporate feedback dynamically and then refine the draft in real time, and that is the reflection agent is significantly improving the comprehensiveness and the quality of the responses.
- Stat: 20% increase in usefulness with Planner Agent
- Credibility: HIGH — Named exec with specific claim and unscripted interview.
- Topic tag:
07-adoption-challenges
Extracted 2026-04-15T00:24:45 via scripts/podcast_mine.py (MLX mlx-community/Qwen2.5-32B-Instruct-4bit).