ChatGPT Agents Are Here—But Are They Ready for the Enterprise?

OpenAI's ChatGPT Agent potentially heats up the already rumbling agent wars that have been ongoing.

However, as these new tools advance, so do the risks. And while OpenAI may have the pure AI muscle to make capabilities like this work, do they have all the additional puzzle pieces that are needed to do so safely?

Microsoft's roadmap for agents (as outlined at Build last year, long before OpenAI's new agent had been revealed) involved enrolling them in identity management provided by Entra.

These Agent IDs will allow agents that are starting to work alongside users instead of at their specific direction to slot into the same security and compliance tooling that organizations using Microsoft 365 have long been used to.

Is this the whole set of puzzle pieces? Certainly not. As I read highlighted by Michael Bargury this morning, the potential problems are only multiplying (post linked in the comments). But at least Microsoft has a set of tools and best practices to leverage against, and broad experience of how to build guardrails that can stand up to the unpredictability of users.

The more tools companies like OpenAI develop, I assume the greater pressure they might feel under to broaden their ancillary capabilities just as we see from Microsoft. But whereas Microsoft's journey has been one of half a century, can OpenAI or the others scale to having such capabilities in just a few years?

Posted on Linkedin 23/07/2025 -> View Linkedin post here