# AI Agents in Legal Tech - David Wong Thomson Reuters on Responsible AI | EP 127

Page: https://stenobird.com/podcast/ai-agents-podcast/ai-agents-in-legal-tech-david-wong-thomson-reuters-on-responsible-ai-ep-127
Text version: https://stenobird.com/podcast/ai-agents-podcast/ai-agents-in-legal-tech-david-wong-thomson-reuters-on-responsible-ai-ep-127.md
Podcast: [AI Agents Podcast](https://stenobird.com/podcast/ai-agents-podcast)
Published: 2026-03-25T15:17:20+00:00
Episode link: https://podcasters.spotify.com/pod/show/ai-agents-podcast/episodes/AI-Agents-in-Legal-Tech---David-Wong-Thomson-Reuters-on-Responsible-AI--EP-127-e3gd1lc
Audio file: https://anchor.fm/s/fe2628e4/podcast/play/116868204/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-2-25%2F420744647-44100-2-5c6db5e4ee4bf.mp3
Processing state: processed
JSON: https://stenobird.com/v1/public/podcasts/ai-agents-podcast/episodes/ai-agents-in-legal-tech-david-wong-thomson-reuters-on-responsible-ai-ep-127
Duration seconds: 2918

## Resource

Thomson Reuters is transforming professional services by moving beyond news into AI-driven software for legal, tax, and audit workflows. The discussion explores how agentic frameworks solve the 'math problem' in tax by using AI to orchestrate specialized tools rather than relying on raw LLM computation.

## Highlights
- Main idea: AI agents in professional services act as orchestrators that use specialized tools (like tax calculators) rather than performing complex math internally
- Practical takeaway: Successful AI implementation requires both 'book smarts' (authoritative reference data) and 'street smarts' (expert human judgment)
- Failure mode: Relying on LLMs to perform direct computation in tax or legal contexts leads to inaccuracies; the solution is API-driven tool use
- Main idea: The Model Context Protocol (MCP) serves as essential 'plumbing' to connect AI agents to disparate professional software and datasets
- Future outlook: AI agents will enable flatter organizational structures, allowing smaller teams to manage high-volume workloads through automated 'army' of agents

## Topics

AI Agents, Legal Tech, Tax Automation, Thomson Reuters, Agentic Workflows, Model Context Protocol, Professional Services, Enterprise AI

## Chapters
- 1:00 — Introduction to David Wong: A brief introduction to David Wong, Chief Product Officer at Thomson Reuters, and his background in B2B software.
- 2:54 — The Thomson Reuters Ecosystem: Understanding how Thomson Reuters operates primarily as a software and research powerhouse for professionals rather than just a news agency.
- 8:40 — Early GPT-3 Testing in Legal: Reflections on the 2020 era of testing GPT-3 for legal research and the realization of AI's potential impact.
- 16:20 — Agentic Solutions for Tax: How agentic frameworks solve the reliability issues in tax preparation by using AI to interface with existing tax engines and APIs.
- 26:55 — Teaching AI to Use Tools: The shift from teaching LLMs to calculate to teaching them how to use calculators and specialized software via APIs.
- 37:45 — The Future of Team Structures: How AI agents will flatten professional hierarchies by enabling smaller, highly efficient teams to handle massive workloads.
- 41:20 — MCP and System Interoperability: The role of the Model Context Protocol (MCP) in expanding the capabilities of AI agents by connecting them to external data and applications.

## Actions

- request_transcript: `POST https://stenobird.com/v1/public/podcasts/ai-agents-podcast/episodes/ai-agents-in-legal-tech-david-wong-thomson-reuters-on-responsible-ai-ep-127/transcription-requests` — Idempotently request low-priority transcript generation for this episode.
- read_markdown: `GET https://stenobird.com/podcast/ai-agents-podcast/ai-agents-in-legal-tech-david-wong-thomson-reuters-on-responsible-ai-ep-127.md` — Read the agent-friendly Markdown representation of this episode resource.

A page view does not enqueue transcription. Agents should invoke `request_transcript` explicitly when they need this episode processed.

## Transcript

Full transcripts are not published on public pages unless there is a clear rights basis.