DevRev, maker of the Computer agentic artificial intelligence product, has launched Voice AI across its Customer Agent for self-service customer support, extending the same organizational memory already powering chat and email into live customer conversations.
By embedding Voice AI into Computer and the agents built on it, DevRev is enabling voice agents to understand enterprise context, reason across business systems, and act while customers are still on the line.
"Voice has become another interface for AI," said Manoj Agarwal, co-founder and president of DevRev, in a statement. "The interesting engineering problem isn't making an agent sound more natural. It's giving an agent enough context to make the right decision while the customer is still on the line. In an enterprise, the answer rarely lives in one place. It lives across tickets, documentation, customer records, orders, engineering systems, and business applications. If an agent can't reason within that context, it must eventually hand the conversation to someone who can. Shared organizational memory changes that by giving an agent enough context to resolve a problem instead of simply responding to it."
Computer, by DevRev, builds a shared organizational memory by continuously synchronizing information across customer records, support tickets, orders, code, documentation, and business applications while preserving existing permissions. With today's announcement, those same agents can now handle live calls.
Because DevRev's Voice AI is grounded in companies' shared organizational memory, it reasons across systems to determine what happened, explains the issue, provides the latest status, initiates the corrective workflow, and documents every action taken, all within a single conversation. When a conversation needs a person, Voice AI hands off to a human agent with the full conversational context intact. Conversations run in multiple languages with the low latency of natural speech, and every call is recorded, transcribed, and available for analysis.
Voice AI in Computer connects to existing enterprise telephony through SIP-based integration. Governance carries over with it, and field-level permissions are inherited directly from connected systems. Every interaction is auditable and every action is traceable.
The architecture behind Voice AI is the same architecture validated through Enterprise-Bench, DevRev’s open benchmark for evaluating enterprise AI systems under real production conditions. Computer achieved 94.3 percent accuracy on the benchmark while using four times fewer tokens per correct answer than Claude Code on identical tasks using the same model, showing that the difference came from system architecture and data context rather than the model itself.