Rakuten(opens in a brand new window) is a world innovation firm working throughout e-commerce, fintech, and cell communications, serving each shoppers and retailers at huge scale. With 30,000 staff worldwide, its engineering groups ship throughout a big, complicated product ecosystem the place each pace and reliability are important.
That’s why Yusuke Kaji, Common Supervisor of AI for Enterprise at Rakuten, has spent the previous 12 months pushing agentic workflows deeper into how groups plan, construct, and validate software program. Codex—the coding agent from OpenAI—has turn out to be a core a part of Rakuten’s engineering stack, particularly the place the corporate wants to maneuver quicker with out compromising safety.
Over the previous 12 months, Rakuten engineers have used Codex throughout operations and software program supply to compress incident response (together with a ~50% discount in imply time to restoration, or MTTR), strengthen CI/CD with automated code evaluation and vulnerability checks, and help extra autonomous growth on complicated tasks.
“We don’t simply care about producing code shortly. We care about delivery safely. Pace with out security will not be success.”
—Yusuke Kaji, Common Supervisor of AI for Enterprise
Inside Rakuten’s engineering staff, their AI agenda is crisp and deliberately operational. Kaji frames the work round three priorities that groups rally behind:
- Construct quicker (“Pace!! Pace!! Pace!!”): Groups use Codex in operational workflows, together with KQL-based monitoring and analysis, to speed up root-cause evaluation and remediation, serving to compress MTTR by as much as 50%.
- Construct safer (“Get issues finished”): Codex is invoked in CI/CD for code evaluation and vulnerability checks, making use of inside requirements mechanically so groups can ship shortly with guardrails.
- Function smarter (“AI-nization”): Codex drives bigger, ambiguous tasks ahead from specification towards working implementations, lowering dependence on perfectly-defined necessities, enabling extra autonomous execution, and finally compressing quarter-long efforts into weeks.
Codex maps straight to every precedence as a reliable agent in a broader toolkit, displaying up the place pace, security, and autonomy create compounding worth.
Pace at Rakuten consists of restoration time, not simply growth velocity.
Groups use KQL (Azure’s question system for logs and telemetry) to observe APIs and analyze indicators. Codex works alongside these workflows to assist establish root causes and counsel fixes, lowering the time between alert and determination.
From a website reliability engineering (SRE) perspective, this shortens the trail from detection to remediation. As a substitute of manually stitching collectively queries, logs, and patches, engineers can give attention to validating and deploying fixes.
Rakuten estimates this strategy can scale back MTTR by roughly 50% when points happen. Or extra merely put: Rakuten has used Codex to repair issues twice as quick when one thing breaks.
As delivery accelerates, evaluation and deployment can turn out to be bottlenecks. Rakuten addresses this by integrating Codex straight in its CI/CD pipeline.
Codex conducts code evaluation and vulnerability checks earlier than modifications attain manufacturing. Rakuten feeds inside coding ideas and requirements into these workflows so opinions align with firm expectations.
“We offer our inside coding ideas to Codex,” Kaji says. “Utilizing the identical ideas, it opinions whether or not the code aligns with our requirements.”
The end result: security checks occur constantly and mechanically, enabling groups to maneuver quicker with out decreasing requirements.
Rakuten’s third precedence—AI-nization—focuses on autonomy. Codex is used not just for evaluation and upkeep, but in addition for executing bigger, ambiguous tasks end-to-end. As a substitute of requiring completely outlined specs, Codex can transfer ahead from partial necessities and produce usable artifacts.
“The most recent Codex fashions can learn between the traces,” Kaji says. “Even when the necessities will not be completely outlined, it understands what we’re making an attempt to construct.”
One instance: constructing a cell app model of an present web-based AI agent service. Codex carried out your complete specification, involving a full stack implementation with a Python/FastAPI backend and a Swift/SwiftUI iOS app, together with all of the backend APIs, with out step-by-step human instruction. Codex minimize the event time for this venture from one quarter to weeks.
As Codex takes on extra code technology work, Rakuten is shifting the engineer’s position to writing clearer specs and verifying outputs towards measurable requirements.
“Our position is to not test each line of code anymore,” Kaji says. “Our position is to outline clearly what we would like and set up tips on how to confirm it.”
Rakuten has supported this shift by means of hands-on workshops throughout engineering, product, and non-technical groups—contributing to Codex taking part in a central position in serving to groups ship quicker, function extra safely, and scale autonomous growth throughout the group.
Be a part of the brand new period of labor
Greater than 1 million companies all over the world are reaching significant outcomes with OpenAI.