November 4, 2026
Startups
DevRev powers Computer and Operator with GPT‑6.1 Sol, striking a balance between reasoning capability, speed, and cost for enterprise services.
Company size: Startup
Region: North America
Industry: Technology
Products: API, Codex
DevRev built Computer and Operator to help service and operations teams resolve even the thorniest customer and employee requests from start to finish. DevRev CTO Ahmed Bashir explains that Computer and Operator are designed to reason across real-time data with reversible actions and fully traceable answers.
DevRev combines OpenAI models—including GPT‑6 Astra and GPT‑6.1 Sol—with a testing harness that connects enterprise knowledge bases and gives teams granular control over agent actions. DevRev found that OpenAI models offer not only exceptional reasoning and reliability when serving customers and employees, but also the requisite speed and cost efficiency.
“It’s an ensemble of capabilities: having just the right amount of reasoning, being trustworthy and well-aligned, and crucially, being viable from a latency perspective.”
— Ahmed Bashir, CTO, DevRev
Grounding AI in enterprise context
OpenAI models provide the core engine behind the reasoning, planning, and tool selection for Computer and Operator. In DevRev’s Agent Studio, users can converse with AI to describe how they want tools to run, while also defining specific tool behaviors through deterministic rules.
DevRev benchmarks these models alongside software responsible for supplying enterprise context, enforcing permissions, and constraining actions. Its Enterprise-Bench benchmark specifically evaluates model performance in complex scenarios where data has multiple versions, information is scattered across disparate systems, and access rights vary by user role.
Ahmed noted that GPT‑6.1 Sol performed exceptionally well in these evaluations. When customers or employees interact with AI agents, the model must grasp the nuances of each request and respond rapidly at a sustainable cost for the enterprise. “I think it strikes that balance perfectly,” Ahmed said.
Building software and sales demos with Codex
Inside DevRev, engineers made a clear choice about the best way to build. “Our entire team uses Codex,” Ahmed said. The tool was adopted organically after engineers experimented with various coding assistants. By the time of the interview, DevRev had fully migrated to Codex, dropping its Cursor licenses.
To Ahmed, this reflects what developers truly value: “Developers are really just looking for the best tool—you can’t fool them.” His engineering team was unequivocal in wanting to stay on Codex.
“We believe the experience Codex delivers is hands down the best.”
— Ahmed Bashir, CTO, DevRev
DevRev’s sales engineers also leverage GPT‑6 Astra to build product demos for the company’s booths at international conferences.
Faster responses to customer and employee needs
When customers or employees reach out for support, a request might require Operator to understand the problem and execute approved actions, or be routed to Computer to assign the right person to take action. Ahmed believes the Decisions API and Ultrafast can help expedite these workflows without requiring deep reasoning at every step.
Some of these decisions are difficult and costly to codify with rigid rules, particularly when the underlying information is constantly shifting. Ahmed wants builders to be able to describe more of this logic conversationally and choose the right form of intelligence for each task. “These operations need to run and finish in seconds, not minutes,” he said.
Delivering on the promise of self-service
DevRev’s next goal is to enable Operator to handle more time-consuming, demanding customer and employee requests—while keeping all actions auditable, permission-governed, and reversible.
In Ahmed’s view, common support tasks like looking up answers in a knowledge base, upgrading plans, or processing refunds are just the beginning. “We want to truly deliver on the promise of self-service,” he said.
Powered by OpenAI’s cutting-edge intelligence, DevRev is advancing Computer and Operator to resolve far more complex requests end-to-end, safely executing actions on behalf of users grounded in enterprise knowledge and permission controls.