How does OpenAI's new deployment strategy differ from traditional consulting partnerships?

DeployCo uses OpenAI's own forward-deployed engineers embedded on-site at customer companies to solve problems and feed real-world challenges back into research. The consulting partners in the Frontier Alliance (Accenture, Capgemini, BCG, and McKinsey) complement this work but struggle to keep up with rapid model changes; OpenAI's partnership ensures consultants deliver the current state of the art to their clients.

Does OpenAI train its models on customer data?

OpenAI does not train on customer data unless a customer explicitly asks for it; those rare cases go through extensive regulatory review on both sides and become formal research partnerships. The feedback loop instead runs through two channels: reports of model weaknesses (such as poor document understanding) that the research team then improves, and tooling needs that feed into products like the open-source Swarm repository and the Agent SDK.

How fast is Codex growing, and which markets lead?

Codex has more than four million weekly users worldwide, a fivefold jump within three months at over 70 percent monthly growth. In Germany, weekly active Codex users have grown more than sevenfold since January 2026, and Germany ranks among the global top five for weekly active users, sits in the global top three for paid subscriptions and developers, and is among the top five for weekly active Codex users.

Back to articles

OpenAI's deployment unit is embedding its own engineers inside major corporations to reshape business processes, not just speed up existing workflows—and customer feedback is flowing back into model development.

THE DECODER4h ago5 min read

Key takeaway

OpenAI has created DeployCo, a subsidiary that embeds its own engineers inside large corporations to co-develop AI solutions, not just implement them. Customer teams identify model weaknesses and infrastructure needs, which flow back into OpenAI's research, creating a feedback loop that has already improved models (such as document understanding at a major bank). Codex, OpenAI's code and task automation tool, now has over four million weekly users globally and has achieved sevenfold growth in Germany since January 2026.

Summaries like this, in your inbox every morning.

3 Key Points

What happened
OpenAI's Deployment Company (DeployCo), unveiled in May with 19 private equity firms, acquired British consulting firm Tomoro, which brings roughly 150 forward-deployed engineers and deployment specialists. Arnaud Fournier, who co-founded OpenAI's Forward Deployed Engineering team, is now CTO of DeployCo. The unit also operates sites in Paris, London, and Munich, and runs joint projects with German firms.
Why it matters
DeployCo's on-site engineers sit at the boundary between OpenAI's product and research teams and customers, carrying lessons back into the company while delivering the latest technology to clients. This feedback loop improves models—for example, document-understanding weaknesses found at major bank BBVA led to improvements from GPT-5.0 to GPT-5.5. The work goes beyond speeding up existing tasks; with BBVA, OpenAI reshaped credit assessment from a once-yearly process into continuous weekly or daily monitoring, so the bank can gauge exposure when geopolitical events occur. For businesses considering OpenAI adoption, this suggests access to engineers who can reshape—not merely optimize—critical processes.
What to watch
Codex (OpenAI's code and task tool) has grown to more than four million weekly users worldwide, a fivefold jump within three months at over 70 percent monthly growth. In Germany, weekly active Codex users have grown more than sevenfold since January 2026, and Germany ranks among the global top five for weekly active users, sits in the global top three for paid subscriptions and developers, and is among the top five for weekly active Codex users.

FAQ

How does OpenAI's new deployment strategy differ from traditional consulting partnerships?: DeployCo uses OpenAI's own forward-deployed engineers embedded on-site at customer companies to solve problems and feed real-world challenges back into research. The consulting partners in the Frontier Alliance (Accenture, Capgemini, BCG, and McKinsey) complement this work but struggle to keep up with rapid model changes; OpenAI's partnership ensures consultants deliver the current state of the art to their clients.
Does OpenAI train its models on customer data?: OpenAI does not train on customer data unless a customer explicitly asks for it; those rare cases go through extensive regulatory review on both sides and become formal research partnerships. The feedback loop instead runs through two channels: reports of model weaknesses (such as poor document understanding) that the research team then improves, and tooling needs that feed into products like the open-source Swarm repository and the Agent SDK.
How fast is Codex growing, and which markets lead?: Codex has more than four million weekly users worldwide, a fivefold jump within three months at over 70 percent monthly growth. In Germany, weekly active Codex users have grown more than sevenfold since January 2026, and Germany ranks among the global top five for weekly active users, sits in the global top three for paid subscriptions and developers, and is among the top five for weekly active Codex users.

Discussion

No discussion yet for this article

Stay ahead with AI news

Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.

Get Started Free

Free · takes 30 seconds · unsubscribe anytime

5 minutes a day. The AI essentials.

200+ sources · Email / LINE / Slack

Get it free →