Anthropic pays $100M to train other companies' engineers: AI's bottleneck is no longer the model, it's who deploys it

By Business Insider · October 12, 2026.

Anthropic this month announced a $100 million investment in a training program for FDEs, short for "Frontier Deployed Engineers": engineers with deep AI expertise and a mandate to create new ways of working inside a company. The company says it wants to train 10,000 people by the end of 2027. The figures come from Business Insider's report, which cites two Anthropic executives.

The format works like a fellowship. Companies nominate their engineers, and Anthropic interviews them. Those accepted spend four days in person building and deploying an agentic system (an AI that carries out multi-step tasks on its own) in a simulated environment. They then take a practical test and return to their company, where they complete a 12-week "residency" with Claude and Anthropic mentors. The first cohorts have started in New York and San Francisco, with engineers from Accenture, Bain and Novo Nordisk, among others. Anthropic itself says it is still building the program's evaluations.

The profile it is looking for combines four traits, according to Shambhavi Shambhavi, head of partnerships strategy: familiarity with agentic systems, problem-solving ability, strong communication and solid engineering skills. Steve Corfield, head of global business development, adds curiosity. Shambhavi stresses that communication matters especially when entire workflows are being redesigned.

The most revealing quote comes from Corfield: "It's a skills gap problem, and that creates a market problem." In other words, customers don't doubt the technology, but they lack people who know how to take it to production.

**Our take.** This is a thesis we have been pointing to for a while: as models become increasingly similar in capability, value shifts to implementation. Anthropic understands this and is funding the scarce link rather than waiting for the market to produce it. It is a smart move, and it cuts both ways.

On the one hand, this is real, valuable training. An engineer who knows how to design, measure and govern an agent within an organization is exactly the profile the transition needs. It is also an example of how work changes rather than disappears: demand is moving toward people who combine technical judgment with the ability to negotiate with people.

On the other hand, it is not philanthropy. The report itself acknowledges that the program aims to get the corporate world used to AI and to increase spending on Claude, just as Anthropic approaches its IPO. Whoever trains engineers on your tool creates habit, standards and dependence. The $100 million is small compared with the recurring spending each cohort could secure. And "gold standard," as Corfield calls it, is a stated aspiration, not a proven result: the program has only just begun and its evaluations are still being built.

In general, companies that take this path will gain speed, but they would do well to ensure their engineers are not tied to a single provider. Other initiatives in the sector (Salesforce, a major Anthropic partner, announced in May a plan to recruit 1,000 junior profiles into "AI builder" roles) suggest training will become a commercial battleground.

In the short term, this widens a gap: large consultancies and pharmaceutical companies will be the first to have these profiles, while SMEs are left waiting. In the long term, if these skills spread, it is a lever for AI to truly reach the real economy, which is the step before its benefits—in health, in productivity—are felt outside the labs.