The coding tool that lived in developers’ terminals now wants a seat in your finance, sales, and design teams.
OpenAI is pushing Codex out of the engineering department and into the rest of the office. On Tuesday it launched six job-specific plug-ins plus new Sites and Annotations features, betting that knowledge workers, now roughly 20% of Codex’s 5 million-plus weekly users and growing three times faster than developers, are the next big market.
Picture the org chart of any mid-sized company. Engineering got the shiny AI tools first. Everyone else watched. Now OpenAI is walking down the hall, knocking on the doors of the analysts, the sales reps, the designers, and the bankers, and saying: this works for you too.
What Happened
On Tuesday, OpenAI rolled out a set of six Codex plug-ins built for specific jobs: data analytics, creative production, sales, product design, equity investing, and investment banking. Each one bundles integrations, instructions, and context so Codex can approximate that role straight out of the box, then sharpen up with customization.
Alongside the plug-ins came an internal report on how Codex is actually being used. The headline number: Codex now has more than 5 million weekly active users, up more than 6x since the desktop app launched in February. Developers are still the biggest group, but knowledge workers now make up about 20% of users and are growing more than three times as fast.
The Backstory
This is OpenAI playing catch-up in a lane it didn’t pioneer. Anthropic launched its Enterprise Agents program back in February, then followed with a finance-focused set of agents in May. OpenAI, with its consumer-first DNA, only brought plug-in support to Codex in March.
So the company known for ChatGPT going viral with the masses is now methodically courting the enterprise buyer it historically left on the table. The timing isn’t subtle either. These features land just three weeks after OpenAI launched the OpenAI Deployment Company, a joint venture for enterprise clients carrying more than $4 billion in funding from global investment firms.
The Core Development
Two more features came with the plug-ins, and they matter more than they sound.
First, Sites. Codex can now output its work as a hosted, interactive website instead of just a local file. OpenAI is launching it with partners including Wix, Base44, Replit, Lovable, Figma, and Emergent, with plans for a bigger partner ecosystem. That turns Codex output from something you save into something you ship.
Second, Annotations. Users can flag a specific part of a document or file, then give Codex more precise commands tied to that exact spot. Less “rewrite the whole thing,” more “fix this paragraph.”
As OpenAI Chief Revenue Officer Denise Dresser framed the broader push, the hard part now is helping companies fold these systems into the infrastructure and workflows that actually run their businesses.
The Business Model Angle
Here’s the pattern worth stealing. OpenAI is running a classic land-and-expand play. Get a beachhead with one user type (developers), prove the tool works, then package the same underlying engine into role-specific bundles to expand the addressable market without rebuilding the product.
The plug-ins aren’t new technology. They’re new packaging. Data analytics, sales, investment banking: same Codex, different presets. That’s the cheat code. Once you have a capable horizontal tool, the growth lever isn’t always more engineering, it’s smart verticalization. You meet each buyer where they already work and hand them something that feels built for them.
For founders, the lesson is sharp: your next revenue line might not require a new product. It might require repackaging the one you have for a customer you’ve been ignoring. OpenAI’s own evolution from research lab to commercial platform is a long study in exactly this kind of expansion, which you can dig into in our OpenAI business model breakdown.
The Risk
Don’t mistake momentum for a moat. “Approximate a specific job” is doing heavy lifting in that pitch. A plug-in that bundles instructions and context is a thin layer, and thin layers are easy for competitors to copy. Anthropic already has finance agents in market. The differentiation here is packaging, and packaging gets commoditized fast.
There’s also the trust problem. A developer can read AI-generated code and catch errors. An investment banker trusting Codex on a deal model, or a sales team leaning on it for forecasts, has higher stakes and less tolerance for a confident wrong answer. “Effective out of the box” is a bold promise in domains where mistakes cost real money. The 6x user growth is impressive, but weekly active users aren’t the same as users who’d cry if you took it away. That’s the metric that actually proves an enterprise play.
Quick Questions
What did OpenAI actually launch for Codex?
Six job-specific plug-ins (data analytics, creative production, sales, product design, equity investing, investment banking), plus a Sites feature that turns work into hosted websites and an Annotations feature for more precise commands.
How many people use Codex now?
More than 5 million weekly active users, up more than 6x since the desktop app launched in February.
Is Codex still just for coders?
Not anymore. Developers are the biggest group, but knowledge workers are about 20% of users and growing more than three times as fast.
Who is OpenAI competing with on enterprise AI?
Most directly Anthropic, which launched its Enterprise Agents program in February and added finance-focused agents in May. OpenAI added Codex plug-in support in March.
The Bottom Line
OpenAI just proved that the smartest growth move isn’t always building something new. It’s repackaging what already works for the people you overlooked. The product is the same engine; the strategy is meeting every role where it sits. For founders, watch whether OpenAI can turn role-specific packaging into genuine stickiness, because in a market where Anthropic copies fast, packaging alone won’t hold the line. Expansion is easy. Retention is the real game.
