News You Can Use

Edition 20 · 15th - 30th June 2025

News You Can Use

Deep Dives

Three stories worth sitting with

AI is breaking the law firm pyramid model

AI is breaking the law firm pyramid model

What
Mark Cohen explains how AI is challenging the traditional law firm model, which relies on a pyramid of junior lawyers doing routine, time-based work. As AI takes over many of these repeatable tasks it reduces the need for large teams of juniors.
So what
This directly relates to our work with our tech stack. As our tools handle more of the routine and junior-level work, we should rethink how junior lawyers are deployed, such as moving them to roles focused on overseeing AI outputs, strategy, and client interactions.

Lexis x Harvey - strategic alliance

Lexis x Harvey - strategic alliance

What
LexisNexis and Harvey have struck a strategic alliance, integrating Lexis's legal databases directly into Harvey's platform. This will allow users to generate answers based on primary law content, with citation-backed responses and custom workflows.
So what
This move reinforces the growing demand for legal AI tools that are deeply connected to trusted content. It is an interesting move from Lexis and leaves the wider market a bit open for others to work with content providers.

A guide to using AI tools properly right now

A guide to using AI tools properly right now

What
Ethan Mollick advises that serious AI users should choose from three leading platforms (ChatGPT, Claude, or Gemini) and emphasises that model selection is key.
So what
This guidance affirms the need for model-tier transparency in our tools. We are testing different models on AGPT for this reason to ensure that we are selecting the right model for different tasks.

Worth Reading

Everything else worth a click

The Leverage - Why Aren't AI Agents Working?

Argues agents face a Catch-22: value grows with freedom, but freedom breaks control via conflicting RLHF and prompt instructions. Until labs solve this, enterprise agents stay as glorified chatbots.

Relativity - Agentic AI Is in the aiR

Relativity's pitch for agentic e-discovery through aiR - autonomous document review and case-strategy work with human oversight. Useful view of how the big e-discovery incumbents are positioning against the new wave.

Ethan Mollick - Using AI Right Now: A Quick Guide

Mollick's practical starter pack: pick one of Claude, Gemini or ChatGPT, pay the $20/month, use the best model and features like Deep Research. Treat it as a thinking partner, not a search engine.

Non-Billable - Inside LegalTechTalk 2025

Harvey at $5B, Legora at $675M, and a clutch of specialist tools (email redlining, AI time tracking, in-house automation) getting real traction. Market is approaching a "commit to your platforms" moment.

Anthropic - Project Vend

Anthropic let Claude Sonnet 3.7 run an automated shop in their office for a month - pricing, inventory, customer interactions. Equal parts promising and chaotic, and a useful ground-truth test on agent autonomy.