OpenAI's GPT-6.1 Sol is already in production at an enterprise software company you've never heard of, handling customer service without human backup.

The Summary

  • DevRev deployed GPT-6.1 Sol to power two autonomous products: Computer (self-service support) and Operator (agent workflow automation)
  • Sol balances reasoning depth, speed, and cost—the enterprise trifecta that determines whether AI agents actually ship or stay in pilot purgatory
  • First real production signal that GPT-6.1's reasoning models can handle unsupervised customer-facing work at scale

The Signal

DevRev, an enterprise service management platform, is running GPT-6.1 Sol in production for two products that operate without human oversight. Computer handles autonomous customer self-service. Operator manages internal agent workflows. Both are live, both are unsupervised, both are handling the kind of work that still employs millions of people.

This matters because it's the first concrete deployment signal for GPT-6.1 Sol outside OpenAI's own products. Sol is OpenAI's reasoning model that sits between o1 and o3. It's optimized for the middle ground enterprises actually need: smart enough to handle complex multi-step problems, fast enough that users don't rage-quit, cheap enough that the unit economics work at scale.

"Balancing reasoning, speed, and cost for enterprise service."

The real tell here is what DevRev didn't build. No human-in-the-loop fallback system. No "escalate to agent" safety valve for the hard cases. Computer and Operator are handling the full stack autonomously. That's not a demo. That's a bet that the model won't catastrophically fail in a way that torches customer relationships.

DevRev's deployment answers the question every enterprise has been asking since o1 launched: can reasoning models handle real customer-facing work, or are they still too slow and expensive for production use. Sol's design explicitly targets this gap:

  • Fast enough for real-time customer service interactions
  • Cheap enough to run on every support ticket without bankrupting the P&L
  • Smart enough to resolve complex issues that currently require L2 or L3 support staff

The Implication

Watch DevRev's customer retention metrics over the next two quarters. If Computer and Operator don't crater satisfaction scores, every enterprise software company will be racing to ship similar autonomous products by Q2 2027. Customer service teams are 5-10% of enterprise headcount. That's the addressable market OpenAI just unlocked with Sol.

If you're building agents or running customer service operations, the new baseline just moved. Autonomous isn't experimental anymore. It's in production, at scale, handling real customers. The question is no longer whether AI can do this work. It's whether you'll be the one deploying it or competing against someone who already did.

Sources

OpenAI Blog