OpenAI Launches DeployCo to Bet Big on Enterprise AI Deployment
Miles Bennett
OpenAI has spun out a standalone subsidiary, DeployCo, and committed $150 million to a consulting-partner network — a clear signal that the AI race is shifting from "who has the best model" to "who can make it work inside a real business."
What is DeployCo, and where did it come from?
DeployCo launched in May 2025, built on two acquisitions: AI consulting firm Tomoro (bringing ~150 engineers) and applied-AI company Northslope, assembling a front-line deployment engineering (FDE) team — engineers who embed inside client companies to make AI actually run — numbering in the hundreds.
The unit raised $4 billion, led by TPG, with Advent, Bain Capital, and Brookfield joining; Goldman Sachs, SoftBank, and Warburg Pincus are founding partners.
This means → OpenAI is no longer just selling model APIs. It is building a turnkey service business — a model company moving onto consulting firms' turf.
What is the real pain point in enterprise AI adoption?
DeployCo CTO Arnaud Fournier put it bluntly: "The gap between model capabilities and what people actually use has never been wider."
In plain terms = large models score well on benchmarks, but companies that buy them often do not know which workflow to plug them into — and when they do, accuracy falls short. The missing piece is not algorithms; it is engineers who understand both the business and the model, on-site, tuning the system.
TPG partner Peter McGoohan estimates the addressable market for AI services will ultimately reach hundreds of billions of dollars — a figure that bets on deployment services, not models themselves.
What does the BBVA case tell us?
Spanish bank BBVA was an early pilot: one OpenAI engineer helped the bank build a credit-risk AI tool, introducing a group of AI agents whose sole job is to review and audit the output of other AI agents. Accuracy rose from below 60% to 80%.
This means → running a model out of the box cannot meet business standards. The critical step is having AI audit AI — dedicated agents verifying another set of agents' results — to push accuracy past the passing line.
This reflects a bigger signal: enterprise AI is not "buy a model and you're done." It is a full stack of engineering plus process redesign. Whoever can deliver that stack is the one who actually captures the revenue.
Why bring consulting firms into the picture?
Gartner analyst Arun Chandrasekaran identified the core gap: AI labs "lack direct access to boardrooms and C-suite conversations" and cannot drive the organizational change that adoption requires.
Boston Consulting, Bain, and Accenture add value not in technology but in change management — persuading departments to cooperate, redesigning workflows, handling personnel transitions.
In plain terms = OpenAI's engineers can make the AI work. Making an entire company willing to use it, able to use it, and committed to using it long-term — that is the consulting firms' job. Two teams, each covering half the problem.
What are competitors doing, and how should we read this race?
The field is already crowded: Anthropic launched its own partner program and Applied AI team in March; Microsoft and Amazon have recently rolled out their own FDE-style initiatives.
DeployCo is still screening its first wave of clients, prioritizing companies with existing long-term OpenAI relationships — locking in reference cases from loyal customers before expanding outward.
This reflects a broader shift in AI competition: once the model layer converges, whoever scales "deployment services" fastest will claim the next phase's profit pool.
Content is for reference only, not financial advice.