Cognizant puts more than 25,000 people into an OpenAI delivery group, with Codex first

Cognizant announced on 30 October an OpenAI Frontier Capability Group of more than 25,000 professionals. The pitch in the release is that model access is no longer the bottleneck for big companies; people who can build with those models inside existing systems are.

What the group will work on

The release names three starting areas:

• AI-native software engineering with Codex. Codex “anchors” this track, helping engineers write, test and ship code.
• Application modernization. Cognizant’s Application Modernizer runs on OpenAI models, and the company says it serves more than 500 mainframe clients, which is the pool this track is aimed at.
• Agentic customer operations. Support and service work handed to agents.

The numbers, and whose they are

Cognizant says more than 21,000 associates have earned over 40,000 OpenAI learning badges, and that more than 12,000 people have taken part in its Codex hackathons. It describes one client project, an OpenAI-powered assistant at a global biopharmaceutical company that it says has handled over 5 million conversations and cut content creation time by 60%. Those are Cognizant’s figures about its own work; the client is not named and there is no outside measurement. Cognizant is one of the Advanced partners in the OpenAI Partner Network.

Where this sits

Systems integrators are picking sides and training at scale. Anthropic put $100 million behind 10,000 deployment engineers earlier this month, and OpenAI’s own enterprise push runs through partners such as Atlassian. The other side of the same trend showed up in Accenture’s reported $29 monthly Claude cap: headcount on a vendor’s programme does not say how much model use each consultant actually gets.

Why this matters

If you buy from a large integrator, expect Codex and OpenAI-based modernization tools to show up in proposals, often with badge counts as the credential. A badge shows someone finished training, not that the code they ship is reviewed differently, so ask how agent-written changes are tested and who owns them. For teams sitting on mainframe code, the modernization track is the part to watch, since that is where Cognizant claims an existing client base.