KB Securities is pushing to commercialise externally its in-house “GgabiAI,” which it has used to improve work efficiency. It is also preparing tokenised securities to offer Korean investment products to wealthy U.S. individuals.

Reuters spoke with Jae-man Park (박재만), managing director and head of KB Securities’ AI and Digital Division, about the firm’s results from applying AI to work, its business-to-business plans and its tokenised securities strategy.

Park, who previously led a big data service development group and a generative AI (Gen AI) group at Samsung Electronics, said AI is already contributing to productivity improvements. He said tokenised securities would be shaped into a business depending on how regulation and the market develop.

AI already in use, IPO demand forecasting cut to 30 minutes

Park said he distinguishes AI as a “certain future” and digital assets as a “future to come” based on actual scope of use and results. He said AI is already being applied to existing work such as information searches, data analysis and document writing, bringing change.

“I described AI as a certain future because it is already contributing and producing results in terms of work productivity,” Park said. He also said the number of users and usage of GgabiAI rose about 20-fold from the start of the year.

In wealth management, incomplete-sales management work fell to 5 minutes from 1 hour. In investment banking, work related to initial public offering demand forecasting was cut to 30 minutes from 7 to 8 hours. In the S&T division’s institutional trading order work, manual tasks fell by about 90 percent.

The AI and Digital Division and the IT division also use AI in development processes. Park described cases of shorter working hours but said it was not easy to convert the firmwide cost-saving effect into an exact amount.

Affiliate rollout from November, push to commercialise GgabiAI externally

Beyond improving work efficiency, the area it aims to link to a new revenue source is GgabiAI’s B2B business. KB Securities plans to broaden its use, starting with rollout to affiliates in November.

“We are currently planning an external business for GgabiAI, in other words a B2B business, and we plan to expand it starting with internal affiliate rollout from November,” Park said.

Asked about concerns that providing it externally could reduce KB Securities’ productivity edge, he said shared AI functions could develop as more client companies join. He said the plan is a structure in which GgabiAI grows through use across multiple companies and businesses, and the benefits are shared.

He added that the current focus remains on strengthening internal competitiveness. “Between external sales revenue and internal competitiveness, internal competitiveness is still far more important,” Park said. “If GgabiAI’s external business expands, external sales revenue could also make a contribution to the company’s earnings that cannot be ignored.”

Shared functions to be common, competitiveness to come from company-specific internal data

Even if GgabiAI is provided externally, not all companies will use the same data and functions. KB Securities is separating data connected to GgabiAI into B2B-provision data and internal-provision data.

It is preparing to allow external client companies to connect their own internal data so they can use a version of GgabiAI tailored to their work. It said common functions can be improved together, but actual work competitiveness can differ depending on each company’s data and how it is used.

“We are preparing so that companies using GgabiAI can connect their own internal data and use a version of GgabiAI specialised for each company,” Park said.

Trust that a client company’s data will not be used elsewhere is also important in the external business. The aim is to manage by separating commonly used functions from company-specific internal information.

AI to assist work, human approval for moving funds

The criteria for applying AI to frontline operations differ depending on the nature of the task. The plan is to deploy AI quickly for internal document drafting and data processing and analysis after completing business tests, then revise and improve it during actual use.

By contrast, he said work involving customer assets or movement of funds requires human approval. He stressed that the role of AI in assisting work should be separated from the role of making final judgments.

“We always assume AI agents are not perfect, and once we finish frontline testing we deploy them quickly,” Park said. “If we improve them through use, agents can be created much faster and fit the field better.”

“When money is moving, I think human approval is essential,” he said, adding that in the long term it could become possible to let AI handle investments or consumption below a certain amount.

Customer touchpoints to shift from broker apps to external AI

He said the expansion of open application programming interfaces (Open APIs) for individuals is likely to affect securities firms’ customer touchpoints. Customers can connect a broker’s services to external AI and use AI to handle information queries, analysis and orders.

He said the process for non-developers to connect APIs directly is not yet convenient, but usage will increase as the user experience continues to improve.

“If a B2B business proceeds so that in ChatGPT you can connect a specific broker’s Open API with one click, the impact on other brokers would be big,” Park said. “That is because the environment of the customer experience is gradually shifting from software apps to AI apps.”

On whether to impose separate control devices for AI orders, he also pointed to difficulty in technical distinctions. If orders come in with the same authentication key, he said, it is hard to tell whether the ordering party is a person, an investment program or AI. He added it is also necessary to examine whether separate controls only for AI match what customers want.

Tokenised securities to review Korean products for wealthy U.S. individuals

The tokenised securities business being pursued with Securitize and Optimism is focusing on wealthy individual investors rather than institutions. The direction is to issue Korean products that U.S. investors may be interested in as tokenised securities.

“The direction we want to pursue with Securitize and Optimism is retail rather than institutions,” Park said. “More precisely, it is retail in the high-asset bracket with a certain level of assets.” He added it would be important to issue Korean products that wealthy Americans may be interested in as tokenised securities.

He said the pace of adoption could differ between AI and tokenised securities. AI can be applied across daily life and work, but tokenised securities can be more affected by regulation and existing systems because they influence specific financial activities such as issuance, trading and transfers.

Customers to split by tech use, human role also important

Park said that over the next 2 to 3 years, different financial experiences for four customer groups could coexist depending on familiarity with AI and tokenised securities.

He said customers using both technologies can collect and analyse information with AI and then move quickly through trading and reinvestment. Customers who prefer existing methods can use face-to-face consultations along with multiple non-face-to-face channels.

Securities firms need to respond to different demands from product design to customer support. He said firms focused on a specific customer group could emerge, but established brokers with broad customer bases and human infrastructure also have opportunities.

Park said adopting AI technology and building customer relationships and networks are separate issues. He said even if AI use spreads, long-built capabilities such as customer response and human networks can remain competitive strengths.

“Because securities firms must serve all four customer groups from product design to customer support, they may need more resources than they do now,” he said. “Securities firms specialised in one side and growing could also emerge.”