[E-Daily Reporter Shin Yeong-bin ] The way global companies adopt artificial intelligence (AI) is rapidly shifting toward a field-based engineering approach. As the ability to successfully integrate AI into actual business systems becomes more important than model performance, the Chonbang Engineering (FDE) model is emerging as the new standard.

Kim Young-wook, Senior Program Manager at SAP France, wrote in a recent opinion piece for Samsung SDS (SAMSUNG SDS CO., LTD.(018260)) Insights: “In just four months, from March to June of this year, AI model companies, software-as-a-service (SaaS) companies, and consulting firms all rallied under the same banner of FDE.” He added, “We are now at a stage where we must view the sales and implementation models of the entire enterprise software industry as undergoing a restructuring.”

FDE is an approach in which engineers go on-site at a client’s premises to understand actual workflows and data environments, and then design and implement AI-based systems right there. The model developed by Palantir for its intelligence agency clients is considered the prototype. It is a structure where engineers go on-site to jointly define and solve problems in situations where clients find it difficult to clearly explain their work processes to external parties or to transfer data outside the organization.

Manager Kim noted that the current environment for AI adoption faced by companies is similar. Companies are unable to clearly define how to integrate AI into their operations, and they also find it difficult to freely provide data to external parties due to concerns regarding personal information, security, and compliance. He explained that, as a result, the ability to successfully integrate AI into actual operational environments has become more important than the AI model itself.

Citing an analysis by the MIT Nanda Initiative, he pointed out that although companies worldwide have invested heavily in generative AI over the past three years, 95% of enterprise AI pilots have failed to generate measurable ROI. He attributed this failure not to insufficient model performance, but to the inability to transition to a system that operates AI within the context of corporate workflows, data governance, legacy systems, and the regulatory environment.

In fact, global companies are rapidly adopting FDE models. Not only large language model (LLM) companies such as OpenAI, Anthropic, and Google, but also SaaS companies like Microsoft, SAP, and Salesforce, as well as global consulting firms such as Accenture and EY, are expanding their FDE organizations or partnerships. Manager Kim observed, “Beyond the observation that model companies are becoming consulting firms, the FDE model is spreading even to the heart of the SaaS sector.”

However, even within the FDE framework, the roles vary by company. The Palantir-style FDE is structured to feed problem-solving experiences gained at customer sites back into the platform’s functionality. LLM companies like OpenAI apply AI directly to customer operations in line with the pace of model evolution, while Salesforce, SAP, and Microsoft integrate AI implementation capabilities onto their existing SaaS platforms. Consulting firms are reorganizing into FDE-style structures based on their existing industry-specific domain knowledge and customer relationships.

This shift holds significant implications for Korea’s IT services and SI industries. Domestic SI companies have accumulated over 30 years of experience in implementing complex systems, including enterprise resource planning (ERP), financial core banking, telecommunications billing, and public administration systems. Global AI models alone cannot quickly grasp the decision-making structures, regulatory environments, and legacy systems of Korean companies, the public sector, and the financial industry. Analysts suggest that this presents an opportunity for the domestic SI industry’s domain expertise to be integrated with the FDE model.

The evaluation criteria of companies commissioning AI adoption are also expected to change. Since the FDE approach involves external engineers accessing a company’s internal systems and data, it has been pointed out that going forward, factors such as data access permissions, security controls, incident response, accountability, and performance-based contract terms—not just performance and price—must be considered together at the Request for Proposal (RFP) stage.

Manager Kim emphasized, “The role of Korean enterprise IT decision-makers is expanding from simply deciding how to adopt global solutions to determining how to integrate global models with Korean domain assets,” adding, “How we leverage the assets of the Korean enterprise IT ecosystem—accumulated over 30 years—will determine our competitiveness in AI.”