The initial value of generative AI was in boosting individual productivity (a 'co-pilot'). The next frontier, and Harvey's current focus, is on building collaborative systems that enhance the productivity of entire teams and organizations, addressing complex problems of orchestration, governance, and workflow.
Harvey demonstrates a successful strategy of targeting a specific, text-heavy vertical (law) with a tailored solution. By deeply understanding the unstructured workflows and data context of legal work, they built a defensible product that generic models couldn't replicate, and are now using that beachhead to expand into adjacent professional services.
Deploying AI in large, traditional enterprises requires more than just a SaaS product; it demands a hands-on approach. Harvey's 'deployed engineering' force and the emergence of law firms as implementation partners show that a services and partnership layer is crucial for integrating AI into bespoke enterprise environments and driving adoption.
The company conceptualizes complex legal matters as reinforcement learning (RL) environments where AI 'agents' can perform multi-step tasks. This forward-looking approach aims to automate sophisticated reasoning by breaking down problems and using expert human feedback (e.g., from senior partners) as the reward signal to improve the system.
Keep pulling the thread on Gabe Pereyra.