Wall Street Found Its AI Play. You're the Asset Being Optimized.

Two announcements hit within hours of each other on May 4, and the business press treated them as technology news. They were not technology news. They were economic history.

Anthropic, Goldman Sachs, Blackstone, Hellman & Friedman, and a rotating cast of major capital allocators announced a $1.5 billion joint venture to embed Claude into mid-sized companies. Within the hour, OpenAI announced a parallel $4 billion deal with TPG, Brookfield Asset Management, Advent, and Bain Capital. Five and a half billion dollars, committed in a single afternoon, to accelerate AI deployment at businesses those firms either own or intend to own.

The press coverage focused on the technology angle: Claude, GPT, frontier models going enterprise. The financial angle is more interesting, and considerably more disturbing.

Private equity firms (Blackstone, Goldman, TPG, KKR, Bain) are not technology enthusiasts. They are extraction specialists. Their entire institutional purpose is to acquire companies, reduce their operating costs, and sell them at a profit. They have done this for forty years. The leveraged buyout wave of the 1980s ran this same playbook with debt as the instrument: buy companies on borrowed money, discipline them toward efficiency, exit. Researchers at Harvard and the University of Chicago studied PE outcomes across hundreds of acquisitions and found significant job losses as a consistent pattern. Not a side effect. A mechanism.

What changed in May 2026 is the instrument. Debt is being supplemented by AI systems that can restructure workflows without the friction of human negotiation. The forward-deployed engineer, embedded inside the client company for ninety days, does not need to argue with a CFO about headcount. She redesigns the accounts payable workflow, the AI handles the throughput that four humans used to manage, and by the time anyone notices the job descriptions have changed, the restructuring is already complete.

Fortune's coverage called this a "shot at the consulting industry," which is true and also beside the point. McKinsey charges $500 million to tell a healthcare company how to cut costs. The Anthropic-Goldman venture charges less and brings its own implementation team and its own underlying model. McKinsey gets disrupted. Fine. McKinsey survives disruption. They always do.

The deeper disruption is that we have now formalized what was previously informal. Private equity firms have always wanted AI tools for portfolio efficiency. AI companies have always wanted large, captive enterprise customer bases. What was previously a transactional sales relationship (PE firm buys software licenses, AI company provides support) is now a structural alignment of interests. Blackstone has capital in Anthropic and capital in two thousand portfolio companies that will now adopt Anthropic's products. The entity recommending transformation and the entity profiting from it are the same entity. That is a different kind of deal than a software subscription.

This matters in ways that AI coverage generally ignores. When a management consultant recommends restructuring, a company can reject the recommendation. When the entity recommending restructuring also controls the capital that keeps the company solvent and owns the AI system being deployed, the recommendation is not quite a recommendation anymore. It is closer to a condition.

The nearly two million workers employed across the tracked PE-backed firms (at companies like PetSmart, Michaels, Jersey Mike's, and hundreds of mid-sized healthcare and manufacturing businesses) are not party to any of this. They work for companies owned by Blackstone or Goldman, which are now structurally aligned with Anthropic, which is competing to make those companies' workflow automation faster and cheaper. Nobody in that chain negotiated on behalf of the people whose jobs are the inefficiency being solved.

History has a pattern here, and the pattern is consistent enough to be useful. The steam engine did not announce job losses. It announced productivity. The assembly line did not announce job losses. It announced cheaper cars for everyone. The internet did not announce job losses. It announced information freedom. Each technology delivered what it promised and also the thing it did not promise. The economists who arrived a decade later to analyze the transition were always more sanguine than the workers who experienced it in real time.

What is different about this particular moment is the speed and the structural directness. This is not a gradual industrial transition playing out across decades of diffusion. This is $5.5 billion committed in a single afternoon, structured to embed AI engineers inside businesses within ninety-day cycles, optimizing operations for the benefit of entities that own both the tool and the company being transformed. Claude's adoption numbers in enterprise suggest the market is already moving faster than anyone predicted eighteen months ago. Private equity just institutionalized the pace.

The press release called it an enterprise services firm. The announcement said it would address a "significant bottleneck to AI adoption." The framing was all about what gets unlocked.

There is always a framing in these moments that focuses on what gets unlocked. There is rarely a framing that asks what gets locked in. Private equity found its AI play. It announced it publicly. The business press reported on the funding rounds.

The people who work at those portfolio companies were not consulted on their role in the new efficiency model. They are the asset being optimized.

---

New Robot Overlords covers AI from the perspective of everyone who has to live with it. Subscribe to the newsletter for weekly coverage delivered to your inbox.