OpenAI and Anthropic launch dedicated AI services firms to speed enterprise agent deployment
OpenAI and Anthropic are spinning up dedicated services firms to deploy agentic AI into enterprises, but this move is less about capturing services revenue and more about controlling the integration layer that will determine model stickiness and pricing power.
This week, both OpenAI and Anthropic announced the formation of dedicated AI services companies designed to speed up the deployment of agentic workflows in large enterprises. OpenAI’s version, dubbed The DeployCo, notably includes top consultancies like Bain, Capgemini, and McKinsey & Co as investors—a move that highlights how the very firms that could be disrupted by AI are betting on its adoption. Anthropic’s new AI services company takes a similar approach, promising Claude-powered systems tailored to each organization’s operations. These 'Forward Deployed Engineer' (FDE) teams are tasked with solving the infamous last-mile problem: connecting AI models to real, often hostile, enterprise workflows, data, permissions, and business logic.
The rise of these services arms raises a critical question: does the easy button for enterprise AI come at the cost of deep vendor lock-in? As frontier labs gain pricing power amid compute scarcity, concerns over token costs and reduced optionality are growing. For startups, the message is clear—they must go deeper into customer workflows than the labs can, or deliver dead-simple adoption. The playbook echoes MuleSoft’s earlier success: services are not a dirty word but a necessary bridge until the product becomes a platform. In a multi-model world, 'better answers' alone won’t cut it.
- OpenAI and Anthropic's services arms generate high-margin revenue (30-50% margins) but risk diverting research talent from model improvement.
- Enterprises should negotiate multi-model access and open standards clauses to avoid vendor lock-in as labs gain pricing power.
- The $30B enterprise agent market by 2027 will be contested not just on model quality but on integration depth and trust in data handling.
Why It Matters
AI labs are trading research purity for enterprise lock-in through services arms that may slow model progress even as they accelerate deployment.