Anthropic Invests $100M to Train 10K AI Engineers
Claude Frontier Academy trains engineers from major firms in a 12-week residency, enhancing competition among AI labs within consulting firms.
Anthropic announced on Friday (December 2) the Claude Frontier Academy, a $100 million program aimed at training 10,000 Frontier Deployed Engineers by the end of 2027. The inaugural class includes engineers from Accenture, Bain, Capgemini, Deloitte, McKinsey, Morgan Stanley, Commonwealth Bank of Australia, and Novo Nordisk, with in-person classes in San Francisco, New York, and London.
The job title speaks volumes. According to Anthropic’s announcement, the Frontier Deployed Engineer leads deployments of agent systems "from faster, redesigned processes to new products and services." The term is derived from forward deployed engineer: an engineer who works within the client’s operation, rather than at the software supplier's factory.
The Format Mirrors Medical Residency
The program has three stages. First, in-person training days with engineers from Anthropic. Next, a simulated deployment at a fictional company. Finally, a 12-week residency where the participant leads a real project with Claude within their original organization. Certification comes at two levels, Claude Resident Engineer and Claude Frontier Deployed Engineer, with the first badges expected to be issued in early 2027.
Steve Corfield, global head of business development and partnerships at Anthropic, summarized the bet: "A small team of highly initiative-driven individuals, with the right skills, access to Claude, and a deep understanding of how the business operates, can transform an entire company." This statement also serves as a marketing argument. Those who train the engineer who designs the process tend to see their chosen model executed.
The program expands the Claude Partner Network, which the company claims has already issued over 175,000 certifications. The residency shifts the focus: it requires a delivered project, not an exam.
Consultancies Trained by the Supplier
The list of participants reveals who Anthropic wants as partners. Of the eight organizations mentioned, five are consultancies that sell precisely the implementation service that the program teaches. Ram Ramalingam, leader of FDE and software engineering and platforms at Accenture, praised the design for testing "judgment against a simulated corporate deployment, not isolated technical skills." At Capgemini, Geoffroy Pajot, global head of AI partnerships, stated that clients do not ask about "the number of certifications or volume of training," but rather how to solve real problems.
There is a tension here that the announcements do not address. Large consultancies maintain partnerships with various labs simultaneously and sell this neutrality to clients. By training thousands of professionals in the methodology of a single supplier, Anthropic pulls this balance to its side. For the buyer, the question becomes whether the integrator recommends the most suitable model or the one in which their team was certified.
The end-user clients bring the most concrete data from the announcement. Rodrigo Castillo, CTO of Commonwealth Bank of Australia, stated that the bank’s teams "produced up to three times more code changes in the past year." This metric measures the volume of changes, not quality or revenue impact, and comes from an institution that is both a client and a showcase for the program.
Where the Effect is Seen Outside the U.S.
The cohorts begin in three English-speaking cities, but the impact spreads across the consultancies' delivery network. In India, Accenture, Capgemini, and Deloitte concentrate much of their engineering centers in Bengaluru, Hyderabad, and Pune. If the FDE model succeeds, the pressure is on the classic arrangement where design stays with the consultant near the client and development goes to the offshore center. A resident engineer who designs and deploys agents within the client shortens this chain.
In Australia, CBA is the only retail bank in the inaugural class, giving it early access to a professional profile that local competitors will still have to train or hire. In Denmark, Novo Nordisk is taking the format to the pharmaceutical industry, a sector where Loic Giraud, vice president of data and AI, stated that they want to provide engineers with "the same practical training" as Anthropic's teams.
The Counterpoint
Training 10,000 engineers in just over a year requires Anthropic to maintain a substantial number of in-house instructors. $100 million divided by the goal amounts to around $10,000 per professional, and the announcement does not indicate whether participants or employers bear any of the costs. There are also, for now, no audited results from completed projects in the residency.
The point of interest for the CIO is another. Until now, the competition among AI labs revolved around benchmarks and price per million tokens. With the Frontier Academy, Anthropic shifts the competition to the client's organizational chart: who holds the resident engineer position determines which model runs in the redesigned processes, and this choice often lasts longer than any benchmark advantage.