Keep pulling the thread on Aatish Nayak, Sachi Shah.
Aatish Nayak advises product teams in the AI space to anticipate the roadmaps of foundation model companies like OpenAI and Anthropic to avoid building features that will be commoditized.
Harvey has found that human preference evaluations from lawyers are the most important metric correlated with customer happiness and retention.
Sierra's business model is outcome-based, where the company charges customers based on the results achieved by its AI agents.
Aatish Nayak predicts that foundation model companies will build native architectures for memory and connectors, making it a risky area for application companies to build their own solutions.
Harvey's customers, including those working with law firm Paul, Weiss, are now demanding "long horizon tasks" like drafting entire 300-page documents.
Harvey works directly with major AI labs during the early stages of model development, receiving approximately one week's notice before a new model is publicly released.
For each turn in a conversation, Sierra's system runs approximately 20 inference calls, including online "supervisor models" that must approve an agent's response before it is sent to the user.
Sierra's live agents use online "supervisor models" that run in parallel with the reasoning layer, preventing a response from being sent unless the supervisor gives a "green light."
Aatish Nayak believes network effects will be a significant competitive advantage for Harvey, as Fortune 500 clients are now mandating that their outside law firms also use Harvey to collaborate on deals.
Harvey's initial go-to-market strategy focused specifically on the transactional corporate legal work sector.
Sierra is focused on building AI agents for the customer experience (CX) market.
Sierra has a long-standing customer relationship with Sonos.