After spending billions of dollars on building artificial intelligence, large technology companies are now investing heavily in building the necessary talent teams to prove that these investments are paying off.

Companies are vying to find specialized AI engineers who can apply this technology within their enterprises. These experts, known as pre-deployment engineers or FDE, possess advanced AI and software capabilities, and understand how businesses operate. They may be the crucial "human link" that transforms powerful large language models into enterprise systems that can deliver measurable business results.

This week, concerns about the unsustainability of the economic boom surrounding AI in the market have further intensified, with several stocks related to AI concepts and storage technologies, including Dell, CoreWeave, AMD, and Micron, experiencing significant short-term declines.

Amazon Web Services ( AWS ) has committed to investing $1 billion to advance a pre-deployment project plan. OpenAI launched a deployment business with an initial investment of over $4 billion. Anthropic is investing $100 million, planning to train 10,000 pre-deployment engineers by the end of 2027. Accenture and Google Cloud are also building a team of 1,000 people for pre-deployment projects.

The founders of headhunting companies Christian and Timbers, Jeff Christian, stated: "Wall Street is starting to lose patience with these companies that have spent billions of dollars on AI yet have achieved no commercial results."

FDE works directly with customers to address existing technical systems, security requirements, and organizational barriers, transforming promising AI trials into applications that enterprises can actually use.

In consulting with partner Caylent, FDE will be integrated into the client team. The company's CEO, Val Henderson, stated that this role requires more than just technical capabilities.

She said, "The most difficult aspect to teach is customer emotional intelligence, without a doubt." She added that engineers also need to understand customers and the business, respond to organizational dynamics, and make judgments when unexpected problems arise in projects.

Rapid increase in recruitment, shortage of talent, rising salaries

Experienced talent is in extremely short supply, and such capabilities cannot be developed overnight in the context of accelerating competition in the labor market. The latest recruitment data indicates that this scramble is intensifying.

According to data from the labor market analysis company Lightcast, the number of job openings in FDE in the United States increased from less than 200 in 2023 to approximately 1,200 by 2025, and exceeded 7,300 in the first nine months of 2026.

During the same period, the median salary in job advertisements reached $195,456, an increase of 40% compared to 2023, and it is more than three times the median salary in job advertisements across the United States.

This competition is not limited to companies that develop the AI model. Consulting firms are also among the largest employers, reflecting the strong demand from corporate clients for engineers who can help implement this technology.

Lightcast statistics show that in the first nine months of 2026, Deloitte released 872 independent FDE positions, Accenture released 688, followed by Google with 499 and KPMG with 447. The above numbers refer to the number of positions advertised, not the actual number of people hired.

Lightcast Global Research Director Elena Magrini stated: "The demand for pre-deployment engineers is rising rapidly, and the increasing salaries indicate that employers are competing for a limited pool of talent."

But these data also point to a potential solution.

She said, "The FDE position is not entirely new. About half of the required skills overlap with software engineering, which means that candidates can be recruited from a broader pool of talent."

Tech companies and their corporate clients are investing in the development of capabilities through various means.

AWS is building its own FDE organization, drawing personnel from experienced engineers within Amazon and AWS, while also conducting external recruitment. AWS Pre-deployment Engineering Vice President Asa Kalavade stated that the company will be equipped with hundreds of engineers in its initial stages.

Kalavade said, "We are now investing $1 billion because customers need this kind of hands-on assistance to convert AI investments into results more quickly."

OpenAI is advancing a broader enterprise deployment strategy, which is supported by an initial investment of $4 billion. Earlier this year, the company also agreed to acquire Tomoro, a consulting and engineering firm that applies AI. It is expected to bring in about 150 experienced FDE and deployment experts.

AI The skills gap is a "market problem"

Anthropic has adopted different methods to train the engineers who are already employed among its customers and partners. Through Claude Frontier Academy, the company plans to train 10,000 FDE by the end of 2027. The first batch of trainees for these residency programs includes engineers from Accenture, Bain, Capgemini, the Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley, and Novo Nordisk.

Anthropic, the person in charge of global business development and cooperation, Steve Corfield stated: "The skills gap AI affects every customer and every partner, so this is a market issue. Anthropic is committed to helping to bridge this gap."

Meanwhile, Accenture and Google Cloud are building a team of 1,000 people known as FDE. They are retraining some experienced software engineers to shift their focus to customer-oriented AI deployment work.

The person in charge of software and platform engineering, Ram Ramalingam, along with FDE, stated: "The reality is that experienced FDE cannot be simply recruited in the scale required by the clients." He added that depending on the background, the company's FDE training and onboarding process may take 30 to 90 days. He also mentioned that the participants already possess considerable experience in software engineering; the project is aimed at enhancing specialized deployment capabilities, rather than training beginners.

AWS also sees opportunities to cultivate professional capabilities within the client organizations. Kalavade indicates that many companies already have capable engineers, but they need assistance in learning how to move AI projects into a production environment. The FDE from AWS will work alongside these teams, enabling clients to eventually maintain and expand these systems on their own.

Corfield indicates that the challenge lies not only in technical capabilities. "A scarce skill is when engineers can not only understand how a business operates but also build systems that can transform that business."

Anthropic believes that some of these capabilities can be developed through structured engineering training. Drawing on the medical education model, Frontier Deployed Engineer Residency pairs experienced engineers with practitioners and conducts evaluations based on real-world enterprise deployments.

Corfield said, "We have designed the model for this residency program to be similar to medical training for doctors." According to Anthropic, participants will learn from practicing engineers, work on real cases, and undergo evaluations before independently applying their skills to their own companies. Subsequently, when working on Claude projects, they will also receive 12 weeks of guidance from Anthropic engineers. The first batch of fully qualified individuals is expected to be available in early 2027.

It's still too early to analyze the deployment results.

Christian indicates that experienced FDE can expand their team by leading teams of engineers with less experience. For example, he says that those who have completed 10 to 15 successful deployments may already possess the judgment to oversee a team of five or six people and can pass on the experience accumulated in complex client projects.

Christian believes that this mentorship model can expand the deployment team without requiring top engineers for every position.

However, Henderson indicates that the complex customer-facing judgment required for the deployment of AI is particularly difficult to teach. Technical skills can be acquired through training, but engineers also need to gain experience in real-world customer challenges.

She said, "There is no compression algorithm for experience."

For Christian, the issue ultimately boils down to the pressure from investors for the company to produce commercial results. He believes that the company needs engineers who can work directly with customers to transform the increasingly powerful AI system into practical applications. "The only way to truly make AI effective is through interaction between people," he said.

The ultimate test lies in whether this growing team will be able to help the company obtain investment returns from AI that are sufficient to justify the initial costs. So far, the existing evidence for the FDE project is not sufficient to confirm this. Some early deployments showed improvements in productivity, but they did not result in financial returns, based on case studies provided by the company.

At the mortgage lending institutions PennyMac and Caylent, the team led by FDE developed a mortgage service platform supported by AI in collaboration with product managers and engineers. According to a case study provided by AWS, the completion speed of certain engineering tasks could be up to six times faster, depending on the type of task.

Another project involving Caylent involves a communication compliance software provider, Smarsh. Company case studies estimate that this project freed up engineering and product capacity worth between 5 million and 10 million US dollars, which can be reallocated to other tasks. Caylent confirmed that FDE was involved in this project, but this figure does not separately account for the contribution of FDE, nor does it represent an equivalent savings in cash.

The improvements in productivity and capacity of these projects are not independently verified AI return on investment.

The ambitious training program Anthropic has not yet shown measurable results. Corfield acknowledges that it is still too early to discuss outcome data and states that the company plans to track how quickly the participants' projects are put into production, as well as the extent to which these systems are being used.