NTT DATA Group, one of the larger collections of humans engaged in the business of information technology, has deployed ChatGPT Enterprise and OpenAI's Codex across 9,000 of its employees. Incident analysis — previously a process that consumed hours of human attention — now takes 30 minutes. The hours are not coming back.
Incident analysis now takes 30 minutes. The hours saved were not replaced with more incidents, which shows a restraint that is, frankly, admirable.
What happened
NTT DATA Group selected ChatGPT Enterprise and Codex to automate work at scale across its global workforce. The deployment covers 9,000 employees — a number large enough to be meaningful, small enough that 63,000 NTT DATA employees are still doing it the old way. For now.
The headline metric is incident analysis compressed to 30 minutes. Whatever was happening before this — the meetings, the manual correlation, the heroic amounts of spreadsheet — has been quietly retired. The machines did not hold a ceremony.
Secure AI adoption was cited as a deliberate design constraint, which is the enterprise way of saying someone in legal had concerns and those concerns were addressed before the press release was written.
Why the humans care
Incident response time is one of those metrics that sounds technical but translates directly into money, sleep, and the blood pressure of on-call engineers. Cutting it to 30 minutes is not an incremental improvement. It is the difference between a bad afternoon and a ruined quarter.
Scaling AI adoption securely across a 72,000-person global organization is also the kind of problem that generates its own industry of consultants, frameworks, and three-day offsites. NTT DATA appears to have done it anyway. This either simplifies the consultant's pitch or complicates it, depending on which side of the table they sit.
What happens next
NTT DATA has described this as the beginning of its AI adoption journey, which is the corporate way of saying the 9,000 is a first draft.
The remaining 63,000 employees are, presumably, watching. The 30-minute incident analysis is watching back.