Japanese IT giant NTT DATA is proving that the era of scaling a business by bloating engineering headcount is officially over. According to an OpenAI case study, the company has shifted critical incident triaging from manual labor to algorithmic rails, slashing analysis time from 15 man-days—where five senior engineers would dig through logs for three days—to a mere 30 minutes of Codex processing. This represents more than just a 99.3% speed increase; it is the systematic elimination of business dependency on scarce technical talent.
Instead of guarding expertise as sacred knowledge, NTT DATA deployed the tool to 9,000 employees, including non-technical staff. Group representatives state that this allows them to delegate clearly defined diagnostic tasks to junior personnel, effectively turning organizational knowledge into executable code. We are witnessing the classic democratization of complex analytics: what once required gray hair and a decade of DevOps experience is now packaged within a ChatGPT Enterprise interface.
The economics of this shift deal a direct blow to the total cost of ownership (TCO) of IT infrastructure. NTT DATA is essentially replacing expensive manual labor with algorithmic anomaly detection. In the world of large-scale consulting, this means complex diagnostic processes no longer need to scale proportionally with payroll. As the company noted, the goal is to empower every employee to transform their role by offloading routine tasks to AI agents.
Key takeaways from the NTT DATA case
Initial diagnostic time dropped 720-fold without hiring additional specialists. 9,000 employees without deep programming skills gained access to complex analytics. AI models are transforming the company’s institutional knowledge into practical tools for frontline staff.
Audit your incident response logs and identify the tasks with the highest diagnostic latency. These workflows are the primary candidates for replacement by agentic models. Otherwise, your budget will continue to leak into "salary bottlenecks" that Codex has already learned to resolve in thirty minutes.