Samsung SDS announces vision to innovate business with generative AI

Hwang Sung-woo, president and chief executive officer at Samsung SDS, delivers a keynote speech at Real Summit 2023 held in Seoul, on Sep. 12. [Courtesy of Samsung SDS]

Samsung SDS Co., the information technology (IT) service unit of South Korea’s Samsung Group, unveiled Tuesday a strategy to target the business-to-business (B2B) market for hyper-scale artificial intelligence (AI) by boosting hyper-automation using existing large language models (LLMs) such as ChatGPT instead of developing its own language model.

Samsung SDS announced its vision for LLM-based generative AI at Real Summit 2023, an event for its enterprise customers, held in Seoul on Tuesday.

“Hyper-automation has evolved for factory automation in areas such as semiconductors, but that for office work has fallen into a slump in the face of difficulties to regularize human speech,” the company said. “LLMs rescued this hyper-automation from the downturn.”

At the event, the company unveiled a generative AI service for B2B customers to help companies automate tasks and maximize productivity. It is based on a multi-engine model that uses various LLMs, including OpenAI Inc.’s GPT-4 and Naver Corp.’s HyperCLOVA X, and combines them with companies’ internal data and assets.

“We are conducting a proof of concept (PoC) to document the work of each employee company wide for AI application,” said Hwang Sung-woo, president and chief executive officer at Samsung SDS.

The company first showcased “Brity Copilot,” a generative AI-based solution that applies to general work systems such as emails, messengers, video conferencing, and data storage.

It records the conversation at a video conference in real time to create minutes or summarize the remarks of individual attendees.

It also acts as a business assistant by guiding the search for relevant internal materials or drafting a report when performing business tasks.

“While Microsoft Corp.’s M365 Copilot supports the same business systems as ours, our solution is unique in that it also supports the payment process,” said Song Hae-goo, executive vice president and head of the Solution Business Division at Samsung SDS.

“We have accumulated much know-how by providing similar services to Samsung affiliates for 20 years.”

Samsung SDS plans to apply its own solution, Brity Copilot, and generative AI to enterprise-wide systems such as enterprise resource planning (ERP), supply chain management (SCM), and human capital management (HCM) by working with partners such as Systems Applications and Products SE (SAP), o9 Solutions, Inc, and Workday, Inc. to design Copilot functions.

“We have quite a few PoCs underway internally, and the development speed increased 30 percent and the verification speed more than doubled after applying one of them to actual development work,” Song said. “As a company that designs and delivers systems to customers, we have been able to significantly reduce the delivery lead time.”

Another solution the company showcased is “FabriX,” a platform that collects and utilizes internal corporate data. It collects numerous data from a company and combines them with LLMs.

With FabriX, employees can easily access in-house assets such as in-house data and create and use custom plug-ins.

“Enterprises will not just use one LLM,” said Koo Hyung-joon, vice president and head of the Cloud Service Business Division at the company. “They can use GPT-3.5 or 4.0, HyperCLOVA X, or even an open-source model. We also provide LLM operations functionality that allows them to select and use these various LLMs.”

Since enterprise solutions utilize internal corporate data, data security is also important.

Samsung SDS says that the solutions run on the Samsung Cloud Platform (SCP), which has enhanced security features for authentication, authorization, logging, and question and answer filtering to prevent leakage of sensitive data.