OpenAI turns ChatGPT into an IT admin

Anyone who has managed enterprise software knows the drill: an employee asks for a quota bump, and you spend ten minutes bouncing across three different dashboards just to check their current usage and hit approve.

OpenAI just launched an Admin plugin for ChatGPT Work and Codex, letting workspace administrators handle all those routine operational tasks directly inside a single chat window.

Admins can now ask questions, dig into workspace analytics, adjust permissions, and approve spending requests without switching tools or writing complex prompts.

Running the workspace through plain text

The plugin covers the core headaches of workspace management: adoption tracking, member provisioning, access controls, and spending limits.

Instead of hunting through reports, an admin can simply ask which teams are hitting their credit limits or who needs access to a specific model. From there, the admin can instruct the plugin to update limits or reassign groups right in the same conversation.

OpenAI made sure to keep guardrails intact. The plugin works strictly within each admin's existing roles and permissions without escalating access. Every action maps directly to authorized system calls, and high-impact changes require explicit confirmation.

It also bridges into everyday communication tools. Admins can route approval requests directly into Slack or Microsoft Teams, letting managers approve or deny extra credits without ever opening the OpenAI admin console.

OpenAI's own IT is ditching manual tickets

OpenAI has been dogfooding these workflows internally to keep its own operations running. According to Kunal Malik, OpenAI's Head of Global IT, the real win is connecting the question to the immediate next action.

Inside OpenAI's own Slack, an internal agent handles incoming IT requests, checks policies, and resolves common issues on the fly.

The company claims this automated setup now resolves roughly 45% of its internal support ticket volume. Even as total IT requests roughly doubled, the team managed to clear out its ticket backlog entirely by turning conversation history into real-time operational dashboards.

Admins can set it up now by enabling plugins in ChatGPT workspace settings and installing the Admin plugin from the internal directory on web or desktop.

Why it matters

We think this is a classic quality-of-life upgrade that hints at where enterprise AI is actually heading: boring, invisible automation that eliminates administrative friction.

For IT managers and small team leads, this is definitely worth turning on immediately. Managing seats, tokens, and budgets across fast-growing teams is usually a mess, and cutting out dashboard juggling will save genuine hours every week.

For the broader industry, it shows OpenAI playing defense against Microsoft. Enterprise buyers do not just buy raw model intelligence; they buy manageable, auditable software. By turning the admin console into natural language tooling, OpenAI is making its standalone enterprise tier much stickier.

The real catch will be trust and precision. Natural language is inherently fuzzier than clicking a radio button in a settings menu. Even with confirmation prompts, trusting an LLM to interpret group-wide permission changes introduces room for human error if an admin phrases a prompt carelessly.