How Ramp engineers use Codex to accelerate code review | OpenAI

May 20, 2026

How Ramp engineers use Codex to accelerate code review

The team is using GPT‑5.5-powered Codex for code review and also building an agent to help manage on-call operations, improving both the developer experience and overall productivity.

Image 1: Ramp customer story artwork for Codex.

Company size: Enterprise

Region: North America

Industry: Technology

Product: Codex

At Ramp, engineers use GPT‑5.5-powered Codex to speed up code reviews and build internal agent tools. That has helped the team get meaningful pull request feedback in minutes instead of hours. Thanks to its reasoning ability, GPT‑5.5-powered Codex can independently reduce much of the manual and human work that would otherwise be required.

“Codex code review finds things that I missed, that another engineer missed, and, of course, things that other AI code review tools would definitely miss.”

— Austin Ray, Ramp AI DevEx

Delivering code review the team can trust

Ramp’s AI developer experience (AI DevEx) team uses Codex to increase the speed of software development and improve code quality.

“Codex code review is the industry gold standard. Ramp has relied on it for years,” explains Austin Ray, who leads AI DevEx. “It’s really good, to the point that engineers ask for it by name. Everyone expects comments on every PR, and it has become essential in many review workflows.”

Before, Ramp engineers sometimes waited hours to get their first review; now they get meaningful feedback from Codex in minutes. What makes Codex stand out is its ability to reason deeply across the entire codebase, bringing what Ray calls “a level of rigor that would be time-prohibitive for many human reviewers.”

In addition to that depth, Codex also offers what Ray describes as “something that engineers can work with the way they work now.” Engineers who prefer to work closer to low-level operations can use the CLI, while the Codex app is available for those who want visual guidance, helpful tools, and other features. Ray himself usually works in the CLI, but the app also appealed to him. “It feels like the app is leading engineering workflows in a more productive direction,” he says.

Video 1

00:00

“GPT-5.5-powered Codex is really good at handling that kind of complexity. If I had to do it all myself, I’d need quite a bit of cognitive load, a lot of sleep, and a lot of focus on the problem.”

— Austin Ray, Ramp AI DevEx

Building internal tools with Codex

Ray also uses Codex to develop “On-Call Assistant,” an agent tool that helps shoulder much of the on-call work Ramp engineers take on in rotation.

“On-call is hard,” Ray explains. “We have a lot of business logic, domain knowledge, and critical incidents. You have to keep a lot of information in context at once and reason through very complex problems.”

For engineers, that isn’t easy. It creates a huge amount of cognitive load and demands sustained focus.

“It’s just complicated,” says Ray. “There are so many concurrency bugs, and the balance between external and internal events is tricky. Plus, there are long-running incident investigations, and the details are always changing, so you have to keep up.”

With Codex, Ray can build with support from its “very strong” reasoning capability. As a result, development of On-Call Assistant has sped up significantly, and Ray feels more confident in the improvements in each release.

“The attack surface of our product is huge,” Ray says. “GPT‑5.5-powered Codex handles it with apparent ease.”

Leadership lessons

Ray is, first and foremost, a platform engineer, and he evaluates all developer tools—including AI-powered ones—from that perspective. In his words: “Does it really change how people ship code, or is it just a demo?”

That is also what Ray recommends to other leaders: prioritize hands-on experience and real-world results.

Show the potential of AI tools in practice: “Have engineers install Codex, sit down with them, and guide them through a strong first experience. Show them what the future of development could look like.”
Create a cycle of trust and improvement: “Most engineers don’t fully understand or believe they can get a good experience from it. They see it as an experimental tool. Walking them through that first experience carefully changes that, gets them to keep trying it, keep improving it, and eventually become one of your best AI users.”
Invest in the feedback loop: “We send feedback directly to the Codex team. If we run into issues, there’s a contact path that gets us there quickly. That feedback loop is where the value of investing in a vendor relationship comes from, and we’ve made incredible progress with the Codex team.”

“Codex is the real deal. Codex absolutely helps us ship faster.”

— Austin Ray, Ramp AI DevEx

Looking ahead

Codex is changing how Ramp engineers work and creating more room to support bigger goals. For Ray, that points to a broader shift in how engineering needs to be approached.

“Engineers are going to become orchestrators. It won’t be about having the ability to write all the code yourself—it will be about knowing when and how to steer AI tools like Codex, when to trust them, and when to push back. At Ramp, the very best engineers are adapting to that shift the fastest.”