Anthropic's $100M Frontier Academy: How the Residency Works

On October 2, 2026, Anthropic launched Claude Frontier Academy, a training programme backed by a $100 million commitment. The target is 10,000 Frontier Deployed Engineers by the end of 2027: engineers who can take a Claude project from an idea to a production system inside their own company. The first programme is a residency, and the only way in is to be nominated by your employer.

Editorial note: Prepared October 3 as an October 2, 2026 dispatch from Anthropic’s announcement and the Frontier Academy programme page, both read on October 3. Anthropic has not published a fee, a curriculum outline or the assessment criteria.

Anthropic’s announcement frames the Academy as an answer to a talent problem. Companies want AI inside their operations, and the people who can make it work there are scarce. Anthropic says a small group of skilled people drives a large share of what AI delivers in the companies it works with, and that the Academy is built from its own deployments inside large enterprises and the consultancies that serve them.

The four headline figures come from the announcement and the programme page. The page describes the $100 million as an initial investment that starts with the residency.

Anthropic compares the Frontier Deployed Engineer Residency to how doctors train: learn from practitioners, work on realistic cases, and pass an assessment before working alone. In practice that means four stages and two badges. The programme page gives the timings that the announcement leaves out.

The design choice that matters is the second half. A certification ends with an exam. Here the exam is the entry ticket, and the credential depends on shipping something real inside the engineer’s own company. One of the partner quotes in the announcement puts the point well.

The announcement lists four traits: hands-on software engineers with strong fundamentals, a record of building with large language models, experience helping other people adopt AI, and a habit of working on their business’s most important problems. Prior experience building AI agents is not required.

There is no open application. Organisations nominate engineers, and each nominee arrives with a named Claude project they will lead when they return. Anthropic tells organisations to ask their Anthropic account team or Partner Account Manager whether they are eligible. That wording points to companies that already work with Anthropic, either as a customer with an account team or as a partner firm. Anthropic has not said whether anyone else can be nominated.

Anthropic names eight organisations in the first cohorts: Accenture, Bain, Capgemini, Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley and Novo Nordisk. Five are consulting firms. Three are enterprises that run their own engineering teams: a bank, an investment bank and a drug maker.

That mix says who the programme is for in its first year. The consultancies will carry residency-trained engineers into client work, which extends Anthropic’s reach well beyond its own staff. The enterprises use it to build internal capacity. Commonwealth Bank of Australia’s chief technology officer is quoted saying its teams produced up to three times more code changes in the past year; that is the bank’s own figure, given in Anthropic’s announcement, with no method attached.

The pattern follows the large consulting commitments we covered earlier this year, including PwC’s plan to train 30,000 professionals on Claude and Cognizant’s 30,000 trained staff. The Academy is a deeper tier for a smaller number of people.

Anthropic now runs three levels of Claude training. The announcement says professionals across 46,000 firms have earned more than 175,000 Claude certifications through the Claude Partner Network, and that nearly 4,000 people have completed Basecamp, which teaches the curriculum Anthropic’s own Applied AI staff receive at onboarding. The residency sits above both.

For agencies weighing which tier matters, our guide to where agencies should certify in the Claude Partner Network covers the services track and its entry requirements. The residency adds a delivery credential on top, not a replacement for it.

The right response depends on whether you can get a place at all. The routes below follow from the eligibility rules, not from any figure Anthropic has published about demand.

The project matters more than the person. A nominee must bring a named Claude project, and the final badge depends on it shipping. A company that picks a vague or low-stakes project wastes its place. A good candidate is one process with a clear owner, data the engineer can reach, and a security review the company already knows how to run.

Most businesses will not get a place, and they do not need one to act. The role itself, which we explain in our guide to the forward deployed engineer, can be hired or grown. For teams that want that capability alongside their own staff, our AI transformation work covers use-case selection, build and handover on the same path the residency teaches.