▲ AI PRISM* Personalized Economic Briefing
* Editor's note: AI PRISM (Personalized Report & Insight Summarizing Media) is an artificial intelligence-based personalized news recommendation and summary service developed with support from the Korea Press Foundation. It selects and provides six news items tailored to each reader type.
[Key Issue Briefing]
■ Insurers with tech DNA: While digital insurers have been absorbed one after another by their parent companies, Kakao Pay Insurance is approaching 10 million cumulative subscribers as it marks its fourth anniversary. Judging that it must win customers on product competitiveness alone, without an agent channel, the company is competing through a developer-centered organization and frequent service upgrades.
■ On-premise AI: Domestic AI ventures are turning to a delivery model that runs AI models on clients' own servers with internet access blocked. Demand for closed networks is also growing among large corporations, public institutions and the defense sector, which worry about leaks of confidential information, and more such attempts are expected.
■ Autonomous manufacturing AI: Manufacturing AI, once limited to detecting defects and anomalies, is expanding into areas where it handles production planning and decisions on equipment operating conditions on its own. Meanwhile, the industry expects demand for technology that reduces reliance on skilled workers to grow quickly as the aging and departure of veteran staff continues.
[News of Interest to Startup Founders]
- Key summary: Jang Young-keun, CEO of Kakao Pay Insurance, presented competition in the manner of a tech company, rather than the conventional grammar of the insurance business, as the way for a digital insurer to survive. Kakao Pay Insurance runs a product-owner (PO) structure in which planning, development, design and insurance specialists form a single team for each product. Developers account for about 60% of total staff, and information technology personnel including planning and design exceed 70%. In the roughly three years since Jang took office, the company has launched some 30 products and made between 200 and 300 service improvements. Its overseas travel insurance grew by word of mouth without large-scale marketing, and the renewal rate has topped 70%. The company has yet to turn a profit, however, and Jang explained that it has moved past the early investment phase into a stretch where revenue growth translates into earnings.
2. Amid Data Leak Concerns, On-Premise AI Running on Companies' Own Servers Gains Ground
- Key summary: Domestic AI ventures are increasingly offering AI models and services in on-premise environments, in which software is installed and managed on a company's own servers. Twelve Labs is discussing plans to install its video analysis AI models Marengo and Pegasus on the own servers of four domestic large corporations and state-run enterprises. If completed, it would be the first on-premise deployment record in Korea. Flitto also won a 3.9 billion won ($2.7 million) contract this month to supply multilingual AI translation and interpretation on an on-premise basis, and MangoBoost agreed to supply its inference optimization software LLM Boost to a defense contractor. Lee Hye-min, head of Twelve Labs Korea, said there is demand to manage sensitive information such as factory secrets and closed-circuit television footage on closed networks rather than external clouds.
3. Beyond Defect Prediction to Process Control: "Autonomous Manufacturing AI" Picks Up Speed
- Key summary: Autonomous manufacturing AI, in which the system decides on its own how best to respond whenever production conditions change, has emerged as a new competitive arena in manufacturing. Nvidia Chief Executive Jensen Huang said in Germany on the 19th that the next wave of AI would be manufacturing, automobiles, heavy industry and robotics, and on the same day Nvidia unveiled autonomous manufacturing technology combined with a digital twin, or virtual factory, at a Siemens electronics plant. In Korea, Ineeji has applied predictive control technology to a heating furnace on KG Steel's continuous galvanizing line, MakinaRocks has applied an AI agent to an electric-vehicle battery production line, and Aizera is demonstrating technology that links quality and process management AI in Pulmuone's tofu process. The industry sees the keys to wider adoption as the work of aligning scattered data from sensors, enterprise resource planning and manufacturing execution systems to a common standard, and safeguards to protect production lines when AI misjudges.
[Reference News for Startup Founders]
4. Finda Wins Approval to Acquire Daewon Savings Bank, a First for a Fintech Firm
- Key summary: Fintech company Finda received approval from the Financial Services Commission on the 21st for the change in major shareholder needed to acquire Daewon Savings Bank. It is the first time a domestic fintech company has acquired a savings bank. A total of 12 billion won ($8.3 million) will be invested in the deal. Finda will acquire a 60% stake to become the largest shareholder, while JB Financial Group will secure the remaining 40% through The Jeonbuk Bank. Daewon Savings Bank, founded in 1988 and headquartered in Gyeongju, North Gyeongsang Province, has effectively suspended operations as its existing shareholders pursued a sale, and its total assets stood at about 3.1 billion won ($2.1 million) as of the first half of this year. Finda's business will accordingly expand from loan comparison and brokerage to the direct supply of financial products such as deposit and withdrawal accounts and mid-rate credit loans.
- Key summary: Toss Securities designed its AI investment information service with the goal of becoming a platform people visit for investment information even when they are not trading stocks. Wang Hyun-min, product owner of the AI tribe at Toss Securities, explained that the service broadened from AI Signal, centered on individual stocks, to Real-Time Issues, which organizes major news, and that on the 14th of this month it added a Market tab covering index and sector moves and foreign and institutional flows. As of the end of September, cumulative users of Real-Time Issues reached 3.12 million, and AI Signal averaged more than 1.5 million monthly users from July to September. Wang said the company treats frequency of information use and revisit rates, rather than trading metrics, as its key performance measures. Toss Securities is also preparing to launch an AI service combining corporate earnings information and earnings calls within the year, and Wang forecast that a hyper-personalization stage would emerge within one to two years.
6. "Claims Paid in One Second — Changing the Conventional Wisdom of Insurance With AI"
- Key summary: Kakao Pay Insurance is running a new experiment in payouts measured in seconds by automating claims assessment with AI. For overseas travel insurance, a system linked to the airport corporation verifies the flight information entered at the time of purchase and pays delay benefits without a separate claim. For mobile phone insurance, the company's proprietary Document AI, which combines optical character recognition with a large language model, reads repair statements or receipts submitted by customers, and about 60% of claims are paid within one minute, with blurry images interpreted from surrounding context. Since the introduction of Forena, an AI search service for employees, the number of calls needed to handle a customer inquiry has fallen from an average of three or four to one or two, and handling time from about 10 minutes to five or six.
