Agentic AI at your complete service: The corporate shift to all-in-one systems

Home > Business > Industry

Agentic AI at your complete service: The corporate shift to all-in-one systems

Published: 14 Oct. 2025, 07:00

LEE JAE-LIM
[email protected]

Audio report: written by reporters, read by AI

Left: Kakao CEO Chung Shin-a speaks at the company's developer conference in Yongin, Gyeonggi, on Sept. 23. Right: Naver CEO Choi Soo-yeon speaks at the company's networking event held in Silicon Valley on June 8. [NEWS1, NAVER]

Once dismissed as a buzzword, agentic AI — artificial intelligence technology that can independently reason and make decisions for users to accomplish their goals — is quickly becoming the next big paradigm shift, one that could redefine how the internet works in the decade ahead.

For Korean companies, this wave represents a rare chance to reset the playing field. Having lagged behind global giants in large language models (LLMs), they are now racing to claim a foothold in agentic AI with sharper, localized strategies. Industry leaders such as Naver — best known for its search portal — and Kakao, the operator of the country’s ubiquitous KakaoTalk messenger, are preparing consumer-facing services to capture users rapidly adapting to the AI era. Mobility players, from automakers to navigation platforms like T Map Mobility, are building conversational copilots reminiscent of Tony Stark’s Jarvis. At the enterprise level, corporations are piloting agentic AI systems that can slash months of cost and effort in business operations.

Together, these efforts mark Korea’s determination to move beyond catch-up in foundational models and carve out a distinct role in the global AI landscape — one where business-to-consumer (B2C), business-to-business (B2B) and mobility services converge to create the foundations of a new economic order.

Agents in daily life
For IT firms like Naver or Kakao, the rise of agentic AI feels like a replay of how portals and messengers once became digital gateways. The race today isn’t just about building smart agents — it’s about owning the distribution, curating them and weaving them into users’ daily routines. Korea’s long-standing dominance in mobile usage and affinity for “super-app” ecosystems give firms a head start if they can make their agent companions feel local, seamless and integrated.

The idea of AI agents embedded in everyday life may still sound futuristic, but Kakao is moving quickly to make it real through KakaoTalk. Kakao’s history is steeped in bold pivots. Founded by Kim Beom-su, the company began as a mobile-first startup in a PC-dominated era. It hit the jackpot with KakaoTalk, which disrupted Korea’s mobile ecosystem by offering unlimited free messaging, later branching into entertainment, mobility and fintech.

Now the company is taking a similar track with AI, using KakaoTalk as the bridge that links users to the wider Kakao ecosystem through agents. Its ultimate goal is to keep users inside the app longer by having AI handle tasks without the need to switch elsewhere. Kakao’s AI assistant Kanana is at the core of this strategy, alongside a partnership with OpenAI that will bring ChatGPT directly into the messenger app.

"Kakao has equipped its agent with ‘Tool Call’ capabilities, allowing it to understand user requests and directly launch the appropriate Kakao services," said Seung Kim, head of Kakao’s AI Synergy Task Force, in a written interview with the Korea JoongAng Daily. "In practice, this means users can seamlessly access various Kakao services — including Gift, Map, Melon, and Talk Calendar — all within KakaoTalk, without having to switch apps or browse through menus."

Seung Kim, Kakao's head of AI synergy task force team [KAKAO]

A real-life scenario the company showcased was a family planning a trip with their dog: Kanana would automatically suggest pet-friendly accommodations and even complete the reservation. Users could then share trip details directly in a Kakao chat, where ChatGPT would coordinate between family members’ schedules and preferences to produce an itinerary. Updates are scheduled to roll out beginning in October.

To power this ecosystem, Kakao is introducing a Model Context Protocol (MCP), a developer framework that simplifies how agents connect across apps. Typically, each new app requires an AI chatbot to learn a separate application programming interface. MCPs replace that with a single standard, so Kanana can call on the right service in response to a user request. Developers can also contribute MCP-compatible tools, which will be uploaded to a new marketplace called PlayTools, allowing users to plug them in like modular apps.

Naver, operator of Korea’s dominant search engine, is following a parallel road map, making search the pillar that connects its services under an agentic AI framework. The company is preparing to transform its core portal into an “AI Tab” by 2027, moving beyond keyword-based queries into chatbot-like interactions similar to ChatGPT. Its pitch is differentiation: drawing on years of accumulated local data from online communities to deliver more culturally and linguistically tailored answers than global rivals.

“Currently, we are developing both the core agent and vertical agents in parallel,” said Kim Sang-bum, head of Naver’s AI Search, in a written interview with the press. “The core agent is responsible for analyzing user intent and planning which subagent to call, while vertical agents are focused on developing the actual functions that connect to real-world actions. Once these two sides are combined, the result will be a single agent where verticals are seamlessly linked at the right moment.”

Unlike Kakao, Naver has yet to partner with a Big Tech firm. Its homegrown HyperClova X remains its publicly known foundational model, though executives have hinted they are “open to other possibilities” for LLM partnerships. Another area of emphasis is multimodality — voice, image, location — as a way to ensure its agents can respond directly to real-world context.

“By integrating multimodal information — voice, image, location — the system can connect directly to real-life actions like ordering, payment or mobility, which we believe will become the foundation for future joint online-offline action-based services that Naver will provide,” the company explained.

Kim Sang-bum, head of Naver's AI Search [NAVER]

Naver is also pursuing a mobility agent through a partnership with Hyundai Motor. By integrating its maps, content, search and scheduling into Hyundai’s upcoming software-defined vehicles, Naver envisions an in-car assistant capable of handling navigation, weather, news, reservations and personal schedules through simple voice commands.

Meanwhile, navigation service operator T Map Mobility, part of SK Square, is advancing its own vision of a full-fledged mobility agent. Leveraging years of local data and dominance in the navigation market, the company plans to merge its services with SK Group’s payments, telecom and electric vehicle infrastructure. The goal is an end-to-end travel orchestrator covering planning, booking, driving, parking and charging.

“We are moving from simple directions to intelligent mobility that understands context and predicts needs,” said Park So-ha, head of T Map’s data and innovation division, at a September press briefing. “Mobility is no longer a single app. It’s an ecosystem in which AI will connect cars, public transit, logistics, and everyday services into one experience.”

Agents in the office
In the enterprise sector, Korean firms are adopting agent platforms that go beyond conventional AI assistants, allowing businesses to build custom AI systems capable of handling more complex, autonomous workflows.

LG CNS, the IT services arm of LG Group, recently unveiled Agentic Works in collaboration with the Canadian AI firm Cohere. The platform lets companies create, monitor and orchestrate their own agents tailored to internal data and business processes. It can be deployed on cloud or on-premises infrastructure, giving flexibility depending on client needs. Alongside it, LG launched AX Sync, a suite of ready-made agents for human resources (HR), scheduling, translation and document management. These are overseen by a coordinating “super agent” that automates daily briefings, approvals, meetings and even live translations. Both platforms are already in use internally at LG Display, where early results show a 10 percent gain in productivity. The company aims for a 30 percent boost within three years, and by swapping foreign subscriptions for AX Sync, it claims to be saving more than 10 billion won ($7 million) annually.

“The future workplace is not humans asking AI, but humans working alongside AI agents that handle entire processes end-to-end,” said LG CNS CEO Hyun Shin-gyoon. “By 2027, we envision a multi-agent ecosystem where specialized AIs collaborate like teams of digital employees.”

Samsung SDS is advancing a comparable strategy. Its FabriX platform lets clients design specialized agents, such as automatic code converters that modernize legacy systems. Brity Copilot functions as a personal assistant, providing translation, resource recommendations, and task execution via voice commands. Meanwhile, Brity Automation applies natural-language prompts to heavy-duty operations in finance, HR and logistics. The company estimates these tools could automate as much as 70 percent of routine office work.

Still, enterprise adoption is in its early stages. According to market tracker Grand View Horizon, Korea’s enterprise agentic AI market stood at just $61.4 million in 2024, but is projected to surge to $875.6 million by 2030. Globally, Gartner forecasts that by 2026, over 40 percent of enterprises will have embedded conversational AI agents, up from under 5 percent in 2023. McKinsey projects that generative and agentic AI could add $2.6 to $4.4 trillion in annual value worldwide, with the largest share coming from automating knowledge work.

Samsung and LG, the largest firms in Korea, are positioning themselves early in this still-nascent field, betting that ownership of the infrastructure — hosting, orchestration, monitoring, governance — will matter as much as the agents themselves. For now, the challenge is to prove that enterprise agents can be reliable, trusted and scalable enough to handle the messy realities of real business operations.

BY LEE JAE-LIM [[email protected]]