OpenAI Codex targets enterprise AI with Sites feature and plugins for white-collar work

OpenAI Codex Shifts Focus to Enterprise AI and Knowledge Work

OpenAI announced a major expansion of its OpenAI Codex platform on Tuesday, transforming the AI coding agent into a comprehensive enterprise AI work platform designed for white-collar work. The update includes three significant capabilities: Sites, a feature enabling users to create hosted interactive web applications; Annotations, an in-place editing tool; and six role-specific plugins that connect 62 popular business applications including Snowflake, Figma, and Salesforce with 110 automated skills built in 2

2

. This strategic pivot positions Codex for enterprise as an orchestration layer for knowledge work rather than solely a developer tool.

The platform now serves more than 5 million weekly active users, representing a sixfold increase since the desktop app launched in February 1

1

. The most striking trend is the user composition: non-developers, including financial analysts, marketers, operations staff, and researchers, now constitute approximately 20% of Codex's user base and are adopting the AI platform for non-developers three times faster than traditional engineers 2

2

. This rapid adoption among knowledge workers signals that AI tools for office workers are expanding beyond the engineering function into broader business roles.

Role-Specific Plugins Connect Business Applications

The six role-specific plugins represent OpenAI's most direct approach to horizontal SaaS integration, bundling connections to 62 business applications with 110 automated skills available immediately 4

4

. Each plugin targets specific job functions: data analytics unifies cloud environments like Snowflake, Databricks Genie, Hex, and Tableau to translate natural language inquiries into reports and dashboards; creative production connects Figma, Canva, Shutterstock, and Picsart to generate campaign assets; sales integrates Salesforce, HubSpot, Slack, and Outreach to automate follow-ups and account reviews; product design bridges Figma and Canva to transform wireframes into clickable prototypes; while equity investing and investment banking plugins sync institutional market feeds including Moody's, FactSet, and PitchBook to streamline financial modeling 4

4

.

These AI tools are designed to be effective out of the box while allowing user customization to enhance performance over time 1

1

. The strategic logic follows a pattern established by competitors: build the agentic AI layer that connects to everything and capture the value of orchestration rather than competing with individual tools 2

2

. A marketing manager who currently switches between Salesforce, Figma, and Snowflake could theoretically manage workflow automation across all three through Codex's natural language interface.

Codex Sites Feature Creates Interactive Enterprise Workspaces

The new Codex Sites feature, launching in preview for business and enterprise customers, allows users to create interactive enterprise workspaces that can be shared via secure workspace URLs rather than producing only local files 1

1

. The practical implication is significant: a financial analyst can take a static spreadsheet, describe requirements in natural language, and Codex will generate a live web application—a scenario planner, dashboard, or interactive model—that colleagues can use without downloading files or navigating spreadsheet tabs 2

2

.

This capability directly threatens the workflow layer currently occupied by tools like Tableau and Power BI, collapsing the gap between wanting an interactive application and having one from weeks of development to minutes of prompting 2

2

. OpenAI is partnering with Wix, Base44, Replit, Lovable, Figma, and Emergent to support the service, with plans to develop a larger partner ecosystem 1

1

.

Annotations Enable Precise Document Editing

The Annotations feature addresses a critical limitation in previous AI document editing workflows. Previously, instructing an AI to update a specific chart or spreadsheet calculation often meant the model had to rewrite the entire file, which frequently broke custom formatting or introduced errors 4

4

. Annotations allow users to designate a specific part of a document or file within Codex, enabling more precise commands and context operations 5

5

.

When a user highlights a specific segment, such as a block of cells in a financial model, Codex isolates those exact data arrays and executes code strictly within that boundary, generating visualizations while leaving surrounding cell dependencies, styles, and unselected formulas completely untouched 4

4

. This localized context-scoping mechanism eliminates full-document regeneration and preserves the integrity of complex business documents.

Growing Use Cases Among Knowledge Workers

The fastest-growing tasks among knowledge workers using Codex include data analysis, up 110% week over week; research, up 37%; and knowledge artifacts—reports, memos, documents, contracts, multimedia assets, PDFs and spreadsheets—up 36% 3

3

. More than 60% of users now run more than one Codex task simultaneously at some point during the day, up from less than half in mid-April 3

3

.

Andrew Hall, a Stanford Graduate School of Business professor, told Axios that he and his students use coding agents like Codex and Claude Code to help with boilerplate academic tasks, data collection, statistical analysis and running code to process data 3

3

. Earlier this year, Hall asked Claude Code to update a paper he'd published five years ago on universal vote by mail, and the tool gathered new data, ran analyses, produced figures and tables and drafted a new paper with minimal prompting. However, when a graduate student audited the work manually, the agent's limits became clear—it made several errors and required expert oversight 3

3

.

Competition with Anthropic Intensifies

The new enterprise features come after a similar push for agentic plugins from Anthropic, which launched its Enterprise Agents program in February, with more specific finance-oriented agents following in May 1

1

. Anthropic's Claude Code and Cowork were the first agentic tools to attract non-coders at scale, with Claude Code launching in October 2025 and going viral after the new year 3

3

. With its traditional consumer focus, OpenAI has been slower to court enterprise customers, only introducing plugin support for Codex in March 1

1

.

The timing of OpenAI's announcement is highly strategic, arriving precisely as its primary investor turned business rival Microsoft kicks off its annual BUILD developer conference in San Francisco, where competing enterprise productivity tools are expected 4

4

.

OpenAI Deployment Company Backs Enterprise Push

The new enterprise features arrive just three weeks after OpenAI launched the OpenAI Deployment Company, a joint venture for enterprise clients that includes more than $4 billion in funding from global investment firms 1

1

. The venture aims to integrate OpenAI tools more deeply into businesses around the world. "AI is becoming capable of doing increasingly meaningful work inside organizations," OpenAI Chief Revenue Officer Denise Dresser said in a statement. "The challenge now is helping companies integrate these systems into the infrastructure and workflows that power their businesses" 1

1

.

OpenAI is attempting to reframe Codex from a tool for developers into something closer to an operating system for knowledge work 3

3

. The 3x adoption rate among non-developers is the statistic that should concern traditional SaaS companies most, suggesting that the market for AI-powered work tools is expanding faster outside the engineering function than within it 2

2

. Whether this orchestration layer can match the reliability, security, and auditability that enterprises require remains the critical question as the preview period unfolds.