The Operating Layer.

OpenAI and Anthropic recently launched deployment companies. $5.5 billion between them. The job: send engineers into client companies and install AI by hand.

This was not the original plan. For three years the dominant theory was that better models would do the work. A company would adopt the model, and the rest would follow. That theory has just been replaced. Both companies have admitted, in public, that the model is no longer the bottleneck. The bottleneck is integration: the data plumbing, the workflow redesign, the change management.

The pattern OpenAI and Anthropic have converged on is the Palantir Forward Deployed Engineer playbook. Embed a senior engineer inside the client company. Learn the workflows. Build the system around the business. Palantir built a multi-billion-dollar company on this model. OpenAI's deployment arm acquired Tomoro, a 150-engineer firm, specifically to get that headcount on day one. Anthropic built its delivery team by embedding applied AI engineers directly inside a $1.5 billion joint venture backed by Blackstone, Goldman Sachs, and Hellman & Friedman.

The unit economics are not subtle. A senior Forward Deployed Engineer costs the firm $250,000 to $500,000 a year, before travel. The math only works at contract values north of a million. That puts deployed engineers within reach of large banks and Fortune 500s. Operating companies below that band cannot reach the price.

Underneath this whole move is a vocabulary the industry is now settling on. The layer that matters is the "operating layer" or the "context layer." Different names for the same idea. It is the surface on top of the model where workflows get redesigned, data gets connected, agents get permissions and audit, and decisions get made legible to software. The model is one part of a much larger system. Most of the value-creating work happens in the operating layer around it.

Most companies are still spending all their AI energy on the model itself. OpenAI and Anthropic have just told you, with $5.5 billion of capital, that the operating layer is what's missing.

This is the work Plainbox Studio has been doing for our clients since we opened. We go deep inside operator-led companies and rebuild the operating layer alongside the team, with the same partner from the first conversation through the last capstone shipped.

If you are reading the deployment moves and wondering what they mean for your company, that is the conversation we have over coffee or by phone, in thirty minutes, with one of the partners.