Here is the core problem that both announcements are responding to. Most businesses that try to adopt AI run a pilot. The pilot looks good in a demo. Then it goes nowhere. MIT's Project NANDA studied this pattern across 300 real deployments in 2025 and found that 95% of enterprise AI pilots deliver no measurable financial impact whatsoever. The AI models are not the issue. The deployment is.
The gap is a knowledge mismatch. Your team knows your business: your data, your compliance rules, your internal systems, your edge cases. The AI company knows how its models behave: what kinds of instructions work, what kinds of data setups produce reliable outputs, where things fail at scale. Neither side has the other's knowledge, and a software subscription cannot bridge that gap. A person has to.
That person is what the industry calls a Forward Deployed Engineer. The term was coined by Palantir in the early 2010s, when it was trying to serve U.S. intelligence agencies whose requirements were too sensitive and too complex to put in a product brief. Palantir's solution was to send engineers to work inside the agencies directly, writing real code inside real systems. Until 2016, Palantir had more of these embedded engineers than it had standard software developers. Critics called the model too expensive to scale. The Q1 2026 results settled that debate: Palantir reported 85% year-over-year revenue growth, with U.S. commercial revenue up 133% in a single quarter.
On May 4, 2026, Anthropic announced a $1.5 billion joint venture with Blackstone, Goldman Sachs, and Hellman & Friedman. The venture will embed Anthropic engineers inside mid-sized companies, initially drawing from the portfolio companies of its investor partners. Apollo, General Atlantic, Sequoia, and Singapore's sovereign wealth fund GIC also backed the deal. The structure gives Anthropic a built-in pipeline of hundreds of companies across industries including healthcare, manufacturing, and financial services.
Seven days later, OpenAI announced something larger. Over $4 billion raised from 19 investors, led by TPG, with Bain Capital, Advent International, and Brookfield as co-leads. The new entity, called The Deployment Company, is majority-owned by OpenAI and valued at $10 billion before the new capital. OpenAI also acquired Tomoro, a 150-person engineering firm that had already built AI systems for clients including Tesco, Virgin Atlantic, and Mattel. The venture is led by OpenAI's chief operating officer, Brad Lightcap.
The investor list tells you exactly who the target customer is. Both ventures pulled in private equity firms with hundreds of portfolio companies: mid-sized businesses in traditional industries that have been trying, and mostly failing, to adopt AI on their own. Goldman Sachs's asset management head described the goal plainly as giving mid-market companies access to engineers they currently cannot afford.
There is also a straightforward competitive story here. Anthropic had been gaining ground in enterprise AI while OpenAI's market share slipped from around 50% in 2023 toward 25% by mid-2025. The Deployment Company is partly a structural response to that shift. Both companies realized that whoever makes it easiest for a real business to go from pilot to production will win the enterprise market.
For traditional consulting firms, this is a serious development. McKinsey, Capgemini, and Bain and Company all invested in OpenAI's venture. That is a hedge: they are backing the entity that could replace them for AI implementation work. The embedded engineer model undercuts the consulting playbook by combining implementation capability with direct ownership of the underlying AI models, faster iteration, and a direct line back to the people who build the technology.
What does this mean for you as a business operator? First, if your AI pilots have stalled, you are not alone and the cause is almost certainly not the AI itself. Second, vendors are now building out services that go far beyond software licensing: engineers who sit with your team, learn your workflows, and build something that actually runs. Third, this will not be cheap. These engagements are aimed at companies that can write substantial contracts, not small operators running a single tool. The mid-market focus, though, means the price point should come down over time as the model scales.
For every dollar companies spend on enterprise software, they spend roughly six on services. OpenAI and Anthropic are moving toward that six.