DevRev’s Use of OpenAI Models
On November 4, 2026, DevRev announced its integration of OpenAI models into its platform to automate customer and employee service tasks. The company utilizes GPT-6.1 Sol and GPT-6 Astra to power two agents, named Computer and Operator, designed to manage requests through to completion while leveraging live company data.
Functionality of Computer and Operator
According to DevRev, Computer and Operator are engineered to reason across live information, generate answers that reference their sources, and perform reversible actions under corporate controls. The system merges model outputs with an execution harness that provides corporate context, enforces permissions, and restricts agent actions.
Agent Studio and Performance Evaluation
In DevRev’s Agent Studio, users can specify how they want a tool to behave in conversations and establish fixed rules for certain behaviors. The company assesses model and system performance through an internal benchmark called Enterprise-Bench, which evaluates the handling of multiple data versions, dispersed information, and role-based access. Ahmed Bashir, Chief Technology Officer at DevRev, noted that GPT-6.1 Sol has performed well in these evaluations.
Development Choices and Tools
DevRev’s engineering teams have standardized on Codex for software development, having discontinued Cursor licenses after exploring various coding tools. Ahmed Bashir stated that the team prefers Codex, with its adoption growing organically. Additionally, sales engineers at DevRev employ GPT-6 Astra to create product demos for trade shows.
Enhancing Workflow Efficiency
For routine decision points that do not necessitate deep reasoning, DevRev highlighted the potential use of the Decisions API and Ultrafast to expedite workflows, aiming for completion in seconds rather than minutes. Ahmed Bashir discussed the critical trade-offs the company considers: reliability, alignment, reasoning capability, latency, and cost, asserting that GPT-6.1 Sol strikes the right balance for enterprise service.
Future Objectives
DevRev’s next goal is for Operator to manage longer and more complex requests while ensuring actions remain traceable, permissioned, and reversible. The company identified common support tasks, such as retrieving knowledge-base answers, upgrading plans, or processing refunds, as initial use cases it plans to expand.
Original source: OpenAI News