Why Enterprise AI Needs Forward Deployed Engineers

News

1. Temasek, Seraphim Back Pixxel in $100 Million Round

2. UN Rights Chief Warns Advanced AI Could Threaten Humanity

3. Tata-owned JLR to Cut About 4,000 Bobs Amid Tariff, China Pressure

4. Anthropic Walks Away From $6 Billion Decart Deal

5. Hyundai India Targets 20% Women in Executive Workforce by 2030

6. Security Breach Drains $320 Million in Bitcoin From Liquid Network

At the end of a 10-month technology overhaul, engineers had less than three hours to move more than 20 million mobile subscribers onto a new cloud platform. Taking the network offline was not an option. A cloud-native telecom software provider supplied the new platform, which replaced several older billing, order-management and customer-facing systems for one of Asia’s largest mobile operators.

Prodapt, a technology services company specializing in telecom , built the migration tools and worked with the software provider and the operator. According to Prodapt, the final switch took under 150 minutes, with no interruption for customers and no loss of data or revenue.

Generative AI automated some of the mapping between the old and new systems. Engineers still had to reconcile customer and product records, connect dependent systems, run repeated tests and prepare to reverse the move if something failed.

Forward deployed engineers, or FDEs, were central to Prodapt’s role. They stayed involved from understanding the operator’s needs and building the tools through the final switch. Prodapt now wants more of its engineers to follow customer projects from the first discussion through development and deployment.

What FDEs do differently

The role predates generative AI. Palantir has long used FDEs to adapt its software to individual customers, while OpenAI and Anthropic now employ them to help customers build and deploy AI applications.

“A conventional delivery engineer builds to a specification someone else has already defined,” says Rajiv Papneja, chief technology officer at Prodapt. “An FDE sits inside the client’s workflow, understands the problem at its source, and owns the outcome.”

When architects, developers and delivery teams each own a separate part, responsibility for the final result can fall between them. AI systems make that problem harder.

Recent MIT Sloan Management Review research points to the same organizational problem. A June 2026 study by Kevin Schmitt, Gregory Vial and Ivo Blohm, based on interviews with 87 practitioners at 23 large organizations, found that leaders creating value with generative AI “expand the scope of use cases across processes” rather than confining the technology to individual tasks.

An AI agent handling a service request may have to read a contract, confirm an account in the billing system, check the customer’s history and record a decision elsewhere. The records may conflict, and each system may have different access rules. The FDE must decide which information the agent can use, what it can change and how errors will be detected.

Pilot projects usually test AI on a narrow task with a controlled set of data. Live systems must work with the records, software and approvals a company already uses. A 2025 MIT NANDA study found that only about 5% of custom enterprise AI tools in its sample had reached production. The authors cited weak integration and poor fit with daily work, but said the number was only an estimate because the organizations studied used different definitions of success.

Training engineers for the role

Prodapt’s FDE enablement program brings together five areas usually divided among several roles: data engineering, full-stack development, machine learning, generative AI, and the monitoring, security and cost control required to run large language models in production.

The company calls this its 5+1 model. The additional element is how the engineer approaches the assignment. FDEs are expected to examine how the customer works, identify who will use the system and understand the result it should improve. In telecom, they also need knowledge of billing, network dependencies, service rules and customer-data controls.

Building that capability required Prodapt to rethink its approach to talent transformation. The company appointed one of its AI delivery specialists to lead the function and introduced a mix of short employee-led sessions, two-day boot camps, online study and mentoring. The shorter sessions have produced more than 90,000 learning hours, while the boot camps have reached more than 900 employees. The 216-hour flipped-classroom program combines self-paced learning, in-person workshops, mentoring and a capstone project, allowing Prodapt to build FDE capability across its global workforce.

“We made Talent Transformation an engine for business transformation,” says Manish Vyas, Prodapt’s managing director and chief executive.

For the final assessment, participants receive a production-style problem and have four to five days to design and build a solution. They must submit an architecture, document their decisions and deliver working code before giving an eight-minute demonstration to Prodapt technical leaders and assessors from its learning partners.

By August 2026, more than 3,500 employees had taken part in at least one element of the program. About 500 had completed the full course and roughly 250 were working as FDEs on customer projects. Prodapt is targeting 2,000 graduates and more than 1,000 deployed FDEs by March 2027.

A common toolkit

Prodapt gives FDEs access to Synapt, its suite of internally developed enterprise AI platforms and tools. Synapt Context Substrate organizes information from databases, documents, policies, manuals and code into a knowledge graph that AI agents can search. Synapt Agent Hub manages guardrails and versions, Synapt Data Transformation supports database migration and modernization, and Synapt Autonomous Operations applies Prodapt’s telecom experience to network operations. 

The context layer helps when enterprise systems use the same term differently. One may treat the bill payer as the customer, another the service user and a third the household. The FDE decides how those records relate, which source takes priority and who may retrieve the information.

“Most AI pilots stall because nobody solves for enterprise context and integration before the demo,” Papneja says. “Both catch up with you the moment you try to scale.”

The telecom migration used Synapt Data Transformation tools. Prodapt says generative AI automated the mapping of older billing and customer-management data to the new platform, reducing manual work by 70%. 

Prodapt had used the same context-grounded GenAI toolkits in an earlier project to migrate and modernize large-scale data workloads on Google Cloud for one of the world’s largest online payments companies. Its case study reported a 50% shorter migration timeline, 30% lower infrastructure costs and 65% better data-pipeline performance. Prodapt adapted that method for the telecom project by adding repeatable tests, audit records and controls for reversing the migration.

Why Japan needs FDEs

In Japan, Prodapt is supplying FDEs to SB OAI Japan, a SoftBank joint venture developing and deploying enterprise AI technology. 

Daichi Nozaki, chief executive of SB OAI Japan, said Japanese companies want to adopt AI but lack enough engineers to introduce it across businesses running decades-old technology. “Right now we need to get the help from outside of Japan,” he told Vyas in an episode of Prodapt’s Alt Shift series.

Nozaki expects FDEs to remain necessary for at least five to 10 years and said Prodapt’s ability to supply them quickly and at scale was one of the reasons for choosing the company.

Prodapt’s own expansion plan is similarly rapid. As the program expands across Europe, North America and Japan, Prodapt will have to maintain the quality of its assessments, give engineers enough industry experience and keep them involved with customers rather than allowing FDE to become another broad job title. 

Prodapt will also have to show that one team’s methods can be used by another, that customers can operate the systems after the engineers leave and that the results promised at the start of a project continue after it enters daily use.

“Enterprise customers are moving beyond AI experimentation and asking what the technology can deliver in practice,” Vyas says. “Our global FDE program and Synapt platforms are designed to turn that ambition into measurable outcomes.”

This partner content article was produced by the MIT Sloan Management Review India Partner Content Team in association with Prodapt.