OpenAI AI Strategy: Codex for Every Job, Hiring More Slowly
How OpenAI uses AI on its own work: Codex as the main tool for almost every employee, internal agents in sales, support and finance, and slower hiring. Updated November 2026.
AI research and products: ChatGPT, Codex, and models and APIs for developers and businesses
OpenAI has published no AI strategy document for its own work. This profile is inferred from its Codex research paper (2026), its OpenAI on OpenAI series (September 2025) and CEO remarks (January 2026).
Read the company sourceWhere OpenAI Is Going with AI
OpenAI is turning its own staff into managers of AI agents. By June 2026, almost all active OpenAI workers used its Codex agent each week. Codex made up 99.8% of the output its workers generated across Codex and ChatGPT, according to OpenAI’s research paper on Codex (company-claimed).
The CEO wants more output from fewer new hires. At a livestreamed town hall in January 2026, CEO Sam Altman said OpenAI would “dramatically slow down” hiring. He added: “So I think the right approach for us will be to hire more slowly but keep hiring” (Futurism, reporting Business Insider).
OpenAI has published no AI strategy for its own work. This profile is inferred from its research, its OpenAI on OpenAI series and CEO remarks. OpenAI sits outside the S&P 100 and STOXX Europe 50, so it is profiled as a Notable company and is not counted in index figures.
The focus has three parts.
- Engineering: Codex for every engineer, making up more than 90% of the average engineer’s AI use by March 2026
- Every other function: legal and recruiting moved most of their AI use to Codex in spring 2026
- Go-to-market and operations: internal agents for sales, support and contract review

Where is your company going with AI, and where will it focus?
Those are the first two of four questions every AI strategy has to answer. Our free guide walks your leadership team through all four in one hour.
What It Takes: Agents in Every Function, No Usage Limits and Slower Hiring
OpenAI puts no limit on how much staff use AI. The Codex paper says there are “no quantity restrictions on AI usage” inside OpenAI, and that staff widely share skills and usage patterns.
Adoption spread from engineers to everyone in 2026. In January 2026, legal and recruiting staff barely used Codex. Their Codex share of AI use reached about 20% by early April and about 75% within a month after that (company-claimed).
Business teams build their own agents. In September 2025, OpenAI launched its OpenAI on OpenAI series with five internal tools: a GTM Assistant in Slack, an Inbound Sales Assistant, a support agent model, a research assistant for support tickets and a contract data agent.
Experts check the agents’ work. Finance experts review every output of the contract data agent. The support team built evaluations so its agents know when not to answer.
Ownership is spread across teams. The tools were built inside go-to-market, user operations and finance. The sources reviewed name no single owner for internal AI use, and Altman sets the stance on hiring.
Headcount is still growing fast. OpenAI had about 4,500 employees in March 2026 and planned to reach about 8,000 by the end of 2026, according to the Financial Times as reported by Engadget. That plan sits uneasily with the CEO’s call to hire more slowly.
AI fluency, coded against the Lead with AI Fluency Matrix (September 2026 edition):
- Fluency floor: Not published. Almost all staff use Codex weekly, but OpenAI has published no requirement or training standard.
- Highest level evidenced: Level 4 (Transforming), company-claimed. Contract review was redesigned around an agent and scaled without matching hires.
- Portfolio design: One agent platform for every function, with business teams building their own tools.

The one-hour AI strategy session
This profile of OpenAI follows the four questions every AI strategy has to answer. Answer them for your own company in one hour.
Our free session guide gives your leadership team the four questions and real answers from across the index.
How OpenAI Measures Results
OpenAI measures AI use in output, and results tool by tool. It reports these figures, all company-claimed:
- Legal and research: the median legal employee generated 13 times more monthly AI output in June 2026 than in November 2025, and the median researcher more than 50 times as much
- Sales: the GTM Assistant gives reps a 20% productivity lift, about one extra day a week with customers
- Inbound leads: the Inbound Sales Assistant raised first-email accuracy from 60% to 98%
- Contracts: the team went from hundreds of contracts a month to more than 1,000, with one new hire
One figure is missing: a company-wide financial value for AI in OpenAI’s own work. Adding one would be the step from AI adoption to AI value.
Next markers: year-end headcount against the 8,000 plan, a value figure, and whether slower hiring shows up in the numbers.
Change log
- September 2025: OpenAI publishes its internal agents for sales, support and contracts.
- January 2026: Altman says OpenAI will dramatically slow down hiring.
- March 2026: The Financial Times reports a plan to reach about 8,000 employees by year-end.
- June 2026: Codex makes up 99.8% of OpenAI workers’ AI output.
The Bottom Line
- Disclosure: 4 of 4. OpenAI answers all four questions through its research and tool write-ups, but publishes no strategy document.
- Every job runs through agents, with Codex as the main tool from engineering to legal.
- Teams build their own tools, with experts reviewing the output.
- The say-do gap is people and jobs. The CEO pledged slower hiring while the headcount plan nearly doubles the company.
- Watch headcount. The test is whether OpenAI grows output faster than its workforce.
Revenue, Headcount and Revenue per Employee
What the numbers show: OpenAI is private and files no audited revenue or headcount figures. It had about 4,500 employees in March 2026 and planned to reach about 8,000 by the end of 2026, according to the Financial Times. These figures do not show that AI caused any change.
Headcount, as reported
- March 2026: about 4,500 employees
- End of 2026 (plan): about 8,000 employees
Notes: OpenAI does not publish audited financial statements, so revenue and revenue per employee are not shown.
These figures give context on scale and productivity. They do not show that AI caused any change. Sources: Financial Times, as reported by Engadget.
How OpenAI Compares with Software and IT Services Peers
OpenAI is a notable company outside the index, profiled for contrast and not counted. Here is how it compares with software and IT services companies in the index.
| Company | Disclosure | AI owner | Fluency floor | Governance | AI value reported | Revenue per employee |
|---|---|---|---|---|---|---|
| Accenture | 3 of 4 | Chief AI and Data Officer | Level 1, training plus tracked use for senior staff | Framework since about 2017; $1B AI safety commitment | Stopped reporting AI sales in 2025 | $91k |
| Adobe | 4 of 4 | No internal AI lead named; finance, IT and security under one leader | Not published | AI Ethics Review Board and impact assessments | Results by team, no total | $758k |
| IBM | 4 of 4 | CEO chairs biweekly AI governance | Not published | Four-layer governance chaired by the CEO | Yes: $4.5B in annualized productivity savings | $256k |
| Intuit | 3 of 4 | Chief AI Officer | Not published; most staff use AI tools | AI Governance Committee, board oversight | Engineering speed and customer outcomes, no total | $1.15M |
| Microsoft | 4 of 4 | Chief Strategy and Transformation Officer | Not published; one AI tool licensed to 200,000+ staff | Responsible AI Standard revised for agents | Results by workflow, no total | $1.49M |
| Oracle | 3 of 4 | Co-CEOs; global CIO runs the internal rollout | Not published; about 80% use AI tools | Standards and security controls before rollout; model costs shown to staff | No: adoption and coding speed only | $478k |
| Palantir | 4 of 4 | CEO | Not published | Not published for internal use | Revenue per employee as the measure | $1.01M |
| Palo Alto Networks | 4 of 4 | Global CIO; VP of Enterprise AI | Not published; AI access under security controls | AI access controls; humans confirm unresolved alerts | Results by process, no total | $524k |
| Salesforce | 4 of 4 | CEO; President of Customer Success runs support agents | Not published | Office of Ethical and Humane Use; agentic AI guidelines | Yes: about $100M a year in support | $498k |
| ServiceNow | 4 of 4 | CEO; Chief People and AI Enablement Officer | Not published | AI Control Tower tracks where AI runs and its impact | Yes: more than $1B in annualized value | $455k |
| Shopify AI Pacesetter | 4 of 4 | CEO sets the AI expectation; engineering leads the tools | Level 2 by expectation: reflexive AI use | AI use rated in performance reviews | Link between AI use and impact, no total | $1.52M |
| Zapier AI Pacesetter | 4 of 4 | CEO leads; AI hackathons for all staff | Level 2 for new hires: Capable or above | Accountability built into the AI fluency rubric | Adoption figures, no value | Not published |
| Anthropic Notable | 4 of 4 | No internal AI owner named | Not published; strong expectation to use Claude | Agent work kept visible; written instructions for agents | Code share and usage at list price, no total | $3.7M |
| OpenAI Notable | 4 of 4 | No single owner; teams build their own agents | Not published; almost all staff use Codex weekly | Experts review agent output | Results by tool, no total | Not published |
Disclosure counts how many of the four strategy questions (where it is going, where it will focus, what it will take and how it will know it is working) a company answers in public. Fluency floor is the lowest level of AI fluency a company expects of all its people, on the five levels of the Lead with AI Fluency Matrix: Avoiding (0), Chatting (1), Collaborating (2), Integrating (3) and Transforming (4). The other columns summarize each company’s profile. Revenue per employee is for each company’s latest fiscal year, in US dollars. Rows marked AI Pacesetter or Notable are companies outside the S&P 100 and STOXX Europe 50. They are shown for contrast and are not counted in any index figure. See all companies in the AI Strategy Index.

Run the same exercise with your own team.
Four questions, one hour, and real answers from the companies in this index, in a free two-page guide.
Frequently Asked Questions
What is OpenAI’s AI strategy for its own work?
OpenAI runs almost all of its own AI use through Codex agents. Business teams build their own agents, and the CEO says AI lets the company hire more slowly.
Do OpenAI employees use Codex?
Almost all active OpenAI workers use Codex each week. By June 2026, it made up 99.8% of their AI output across Codex and ChatGPT, according to OpenAI.
Is OpenAI cutting jobs because of AI?
No. Altman said in January 2026 that OpenAI would hire more slowly but keep hiring, and a March 2026 report put its plan at about 8,000 staff by year-end.
What internal AI tools has OpenAI built?
OpenAI has published five internal tools: a GTM Assistant, an Inbound Sales Assistant, a support agent model, a research assistant for support tickets and a contract data agent.
How is this profile put together?
Lead with AI profiles use public sources only, such as annual reports, earnings materials, filings and on-record interviews, dated at the time of writing. AI fluency is coded against the Lead with AI Fluency Matrix, and company figures are marked as company-claimed unless they have been independently verified.
About the AI Strategy Index
The Lead with AI AI Strategy Index profiles how the largest companies in the US and Europe approach AI, starting with the S&P 100 and the STOXX Europe 50. Each profile answers four questions: where the company is going with AI, where it will focus, what it will take and how it will know it is working. AI Pacesetters and notable companies outside the two indexes are profiled for contrast and are not counted. Profiles use public sources only, and AI fluency is coded against the Lead with AI Fluency Matrix.
See all company profiles