AI agents are getting smarter fast, but their token costs may be heading toward a pricing reckoning. In this condensed recap of Sourcery, host Molly O'Shea talks with Harvey co-founder Gabe Pereyra and Niko Grupen about why model performance, cost, latency, and trust now matter as much as raw capability. Instead of the full episode, you can get the key ideas in minutes: how Harvey’s open-sourced Lab benchmark evaluates legal AI on real tasks, why “legal unit tests” make sense for work like due diligence and contract review, and why model routing is becoming essential as different models win on different price points. The discussion also covers the shift from chat copilots to agentic systems, the infrastructure needed for sensitive legal data, and how companies can build organizational intelligence on their own data. Listen now to get the key ideas in minutes.
Podden och tillhörande omslagsbild på den här sidan tillhör
Transcripted.ai. Innehållet i podden är skapat av Transcripted.ai och inte av,
eller tillsammans med, Poddtoppen.