TL;DR
OpenAI and Anthropic announced parallel enterprise deployment moves in early May 2026, according to the source material, placing engineers inside companies to turn AI pilots into production systems. The bet is that services, not model access alone, will decide the next phase of enterprise AI adoption.
OpenAI and Anthropic moved into the enterprise services layer in early May 2026, according to the source material, launching parallel structures meant to place engineers inside companies and turn AI systems from pilots into production tools. The move matters because it shifts the contest among AI labs from selling model access alone toward controlling deployment, integration and long-term enterprise usage.
Anthropic announced a $1.5 billion enterprise-services venture with Blackstone, Hellman & Friedman and Goldman Sachs to embed Claude inside mid-market companies, according to the source material. Hours later, OpenAI announced a $4 billion Deployment Company, described as “DeployCo,” at a $10 billion pre-money valuation, with 19 investment partners.
OpenAI also acquired consulting firm Tomoro, bringing in 150 forward-deployed engineers on day one, according to the same material. The stated model is patterned on Palantir: engineers work on-site or closely with client teams, learn operational workflows, build software around a frontier model and remain involved until the system is running in production.
The core claim behind both moves is that model capability is no longer the main barrier for many enterprise customers. The source material says the bottleneck has shifted to integration, security review, evaluation systems and redesigning business processes around AI.
Why It Matters
The development signals that major AI labs are no longer treating services as a side channel. They are moving into the work that consultancies and systems integrators have long monetized: implementation, workflow redesign and operational change.
The economic reason, according to the source material, is the ratio between software and services spending. For every dollar companies spend on software, they spend roughly six on services. If that ratio holds in enterprise AI, the larger revenue pool may sit not in model subscriptions alone but in getting those models embedded into daily work.
The strategy could also deepen customer lock-in. Once an AI system is built around a company’s internal processes, data flows, evaluation methods and operator habits, switching providers may become harder. In a token-metered model, revenue can also grow as the deployed AI system does more work inside the customer’s business.
Building Generative AI Services with FastAPI: A Practical Approach to Developing Context-Rich Generative AI Applications
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
Palantir developed the forward-deployed engineer model through years of defense, intelligence and enterprise work. In that model, the engineer is not only a consultant producing recommendations; the role is tied to building and adapting software for a specific operational problem.
The source material frames OpenAI’s and Anthropic’s moves as a response to stalled enterprise AI adoption. It cites MIT research saying 95% of generative-AI pilots fail to move beyond the experimental phase. That figure is used in the material to support the argument that many companies can test AI tools but struggle to run them at scale.
The analysis also links the move to pressure on the model layer. If frontier models become harder to differentiate over time, the labs may seek advantage through deployment capacity, customer integration and usage-based expansion.
“DeployCo”
— Source material on OpenAI’s new entity
“almost line for line”
— Source material on the deployment structure
“model performance is no longer the constraint”
— Source material on enterprise AI adoption
“resembles consulting more than pure software licensing”
— Source material on the risk profile
AI Augmented Consulting: Master Prompt Engineering to Build Authority, Generate Leads, and Deliver Client Insights (AI in Action: Practical Artificial Intelligence for Every Industry)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
Several points remain unclear from the source material. It does not establish how fast either deployment structure can scale, how margins will compare with software licensing, or how much customer demand will convert into long-term production usage.
The largest open question is whether forward-deployed engineering becomes a product-formation engine or a labor-heavy services business. If each customer requires proportional engineering time, the model may pressure margins. If repeated deployments produce reusable systems, the labs could turn services work into a stronger software platform.
Redesigning Workflows for AI: How to Transform Processes, Decisions, and Organizations in the Age of Intelligent Systems
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
The next test is execution: how many enterprise customers move from pilot projects into production deployments, how much recurring usage those systems generate and whether the labs can standardize what their embedded engineers build. Investors and customers will also watch whether OpenAI’s Tomoro acquisition and Anthropic’s services venture produce measurable adoption beyond initial announcements.
AI Programming Made Practical: A Step-by-Step Guide to Building AI-Powered Applications, Writing Better Code Faster, and Using Modern AI Tools with Confidence
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What happened?
OpenAI and Anthropic announced enterprise deployment moves within roughly 72 hours in early May 2026, according to the source material. Both are aimed at embedding engineers with companies to put AI systems into production.
Why are AI labs moving into services?
The source material says enterprise AI adoption is being held back by integration, security review, evaluations and workflow redesign. Services address those barriers more directly than model access alone.
What is a forward-deployed engineer?
It is an engineer who works closely with a customer’s operators, learns the workflow and builds software around the customer’s specific problem. The model is associated with Palantir.
What is the main risk?
The model may be labor-intensive. If every new customer needs heavy engineering support, margins could come under pressure. If the work produces reusable product patterns, the model could scale better.
Source: Thorsten Meyer AI
