We Built Our Own Cloud Agent Infrastructure (www.harvey.ai)

The company has transitioned its AI product, Harvey, from a chat service to a fully-fledged cloud agent that can perform complex legal tasks, such as reviewing large volumes of documents. Recognizing that existing managed agent infrastructures from major players like Anthropic and OpenAI did not meet the specific demands of law firms, particularly around multi-model flexibility, zero data retention, and cost control, they opted to build their own cloud agent infrastructure. This decision allows them to navigate conflicts of interest, ensure client confidentiality, and maintain economical operations tailored to the legal sector's unique requirements. The significance of this development lies in the growing necessity for law firms to leverage various AI models without being locked into a single provider, as client confidentiality and operational effectiveness increasingly demand versatility. The firm’s infrastructure features an abstraction layer that standardizes differing agent harnesses, enabling seamless model routing to optimize performance and control costs. This strategic investment not only meets immediate client needs but positions them to adapt as the cloud ecosystem evolves—ultimately allowing greater customization and jurisdiction-specific compliance while serving the expanding requirements of regulated enterprises.