NTT DATA cuts incident response time from 72 hours to 30 minutes using OpenAI's Codex

NTT DATA’s agentic AI breakthrough cuts incident analysis from days to minutes

NTT DATA Group, a Japan-based global IT services provider, deployed OpenAI’s Codex to approximately 9,000 employees and achieved a dramatic productivity gain: automating a critical system incident analysis that previously consumed three days and five engineers into a 30-minute automated workflow. According to the OpenAI Blog, the success signals a shift in how enterprises view AI—moving beyond conversational assistants to autonomous task execution. The deployment builds on NTT DATA’s May 2025 strategic partnership with OpenAI and company-wide rollout of ChatGPT Enterprise, which achieved 96% user satisfaction and 95% reported productivity gains.

From ChatGPT conversations to agentic automation

NTT DATA established an internal OpenAI Center of Excellence to support adoption of ChatGPT Enterprise, distributing licenses, validating technical implementations, and developing use cases across the organization. According to the OpenAI Blog, employees initially used the platform for research, writing, and content creation—building familiarity with AI reasoning before graduating to task automation. This sequential adoption strategy—conversation-first, then automation—created organizational readiness for Codex’s more autonomous capabilities.

The 30-minute incident response that proved agentic potential

The breakthrough came when NTT DATA’s technical teams applied Codex to a longstanding bottleneck: diagnosing failures in a critical system. According to the OpenAI Blog, the manual process had required five experienced engineers and 72 hours. Codex independently executed the full diagnostic pipeline—investigation, execution, testing, and iterative revision—in half an hour. Hiroaki Sato, an AI Technology Department official at NTT DATA Group, told the OpenAI Blog that the result fundamentally shifted internal perception: “Codex has changed how people across the company think about AI. The idea that AI can take the lead in carrying out work has had an impact similar to the arrival of ChatGPT.”

The incident analysis use case proved to leadership that Codex delivers value beyond code acceleration. The speed-up—from 4,320 minutes to 30 minutes—removed a significant constraint on production reliability and freed senior engineers for higher-level design work. According to the OpenAI Blog, the momentum from this win prompted wider exploration: teams across technical and nontechnical roles began identifying new use cases for agentic task execution.

Why this matters

For enterprises evaluating agentic AI: NTT DATA’s 30-minute incident diagnosis demonstrates that autonomous systems can handle complex, multi-step workflows that historically required specialized human judgment. This reframes the ROI calculation—instead of asking whether AI can assist knowledge workers, organizations can now ask which time-bound, clearly scoped processes can be fully delegated to AI agents. For IT operations teams specifically, incident response automation at this speed could fundamentally change on-call rotation economics and system reliability SLAs.

For OpenAI’s positioning: The case study shows Codex penetrating nontechnical workflows at a 9,000-person scale, validating the bet that agentic capabilities extend beyond software engineering. If reproducible across other domains, this signals a market shift from AI-as-assistant to AI-as-performer, likely accelerating enterprise licensing and justifying higher per-seat pricing for ChatGPT Enterprise and specialized agentic products.

For incident management vendors: The automation threat is no longer hypothetical. If Codex can compress three-day diagnostics into 30 minutes, legacy ticketing and incident management platforms face pressure to embed or partner with agentic AI, or risk becoming annotation layers on top of AI systems rather than decision-making tools.