Samsung just pulled back the curtain on what it took to transform Bixby from a voice assistant into a full-fledged AI agent. In an exclusive interview, Jisun Park, Corporate EVP and Head of Language AI at Samsung's Mobile eXperience Business, revealed the company completely rebuilt Bixby's architecture around large language models - ditching the old command-based system for one that can actually reason, plan and execute complex tasks autonomously. The new Bixby launched March 31, and it's Samsung's bet on the agentic AI future.
Samsung isn't playing around with voice assistants anymore. The company'sLanguageAI chief just laid out exactly how they transformed Bixby into what they're calling a "device agent" - and the technical details reveal a fundamental architectural overhaul that puts it squarely in the agentic AI race.
Jisun Park, who runs Samsung's Language AI Team, told Samsung Newsroom the biggest challenge wasn't adding new features. It was "redesigning Bixby's architecture from command-based to agentic." That's a massive shift. The old Bixby classified user input and executed preset scenarios. The new one, powered by an LLM core, interprets intent and generates its own execution plans on the fly.
Here's how it works now: Samsung transformed individual device functions into callable agents that the LLM can invoke as needed. "We defined them in a way that allows the LLM to invoke them as needed," Park explained. "This enables the system to combine multiple functions and APIs to complete tasks more meaningfully." Instead of mapping "turn on dark mode" to a specific command, Bixby now understands when someone says "my eyes are tired" and reasons its way to activating Eye Comfort Shield.
The natural language understanding got a serious upgrade too. Users can ask vague questions like "make my screen visible only to me" and Bixby activates Privacy Display. No menu diving, no exact feature names required. It's understanding intent, not keywords. Park says this makes it "essentially a service center in your pocket" - the device troubleshoots itself based on context and current settings.
But the real technical achievement? Korean language processing. Park called it "particularly memorable." Korean's extensive use of particles and verb endings creates massive variation in word forms. Flexible word order and heavy context-dependence mean the same sentence structure can carry wildly different meanings. "These characteristics make it challenging for LLMs to reliably interpret sentence structure and semantics," she said.
The team refined their LLM training approach, improving model architecture and strengthening context-based learning. The result? Korean language performance that "well beyond our initial targets." That breakthrough gave the team confidence the entire agentic architecture would work. It's a reminder that building multilingual AI agents isn't just about translation - it's about fundamentally different linguistic logic.
The new Bixby also breaks out of the device silo. It now pulls real-time web information directly into conversations. Ask "recommend three Korean restaurants in Seoul for a family of four" and you get results without opening a browser. Follow-up questions flow naturally. Samsung's integrating search context into the agent layer, keeping users in a single conversational flow instead of bouncing between apps.
SmartThings integration extends this further. Through Samsung's smart home platform, users can control appliances remotely via Galaxy devices. "Start cleaning the floor" activates your robot vacuum. "Turn on the air conditioner in dehumidification mode" preps your home before you arrive. Park frames this as Bixby becoming "the primary entry point for interacting with Samsung products."
That's the real play here. Samsung wants to kill the app-centric interface model. "In the past, users had to search for the right app, navigate menus and move between multiple screens to complete a task," Park said. "With Bixby, simply speaking is enough." It's a shift from tap-and-swipe to pure conversation - and it requires AI that can actually reason about what you want, not just execute preset commands.
The timing matters. OpenAI, Google, and Apple are all racing toward agentic AI assistants that can take actions across apps and services. Samsung's advantage is vertical integration. They control the hardware, the OS, and now an LLM-powered agent that deeply understands device state. Park emphasized Bixby is "optimized for each user's device" and "deeply understands device status and capabilities."
Rolling out across the Galaxy S26 lineup and expanding to more Samsung devices in phases, Bixby 4.0 represents Samsung's architectural answer to the agent era. The company's betting users will adopt AI faster if it's embedded in devices they already own, doing tasks they already need done - no technical expertise required.
Park's vision is ambitious: Bixby as infrastructure, not just a feature. "Samsung aims to accelerate the widespread adoption of AI and ultimately embed it seamlessly into everyday life, much like essential infrastructure," she said. Whether users actually talk to their devices instead of tapping apps remains the billion-dollar question. But Samsung just showed it's rebuilt the entire foundation to find out.
Samsung's Bixby overhaul is a technical statement about the future of device interfaces. By rebuilding the entire architecture around LLMs and agent-based reasoning, Samsung is positioning itself against the coming wave of agentic AI from big tech rivals. The real test won't be whether Bixby can understand Korean verb conjugations or activate privacy modes - it'll be whether users actually shift from tapping apps to talking to their phones. But with deep device integration, SmartThings connectivity, and an LLM that can reason through complex multi-step tasks, Samsung just put serious infrastructure behind the conversational computing bet. The company controls the full stack from silicon to software, and now it's got an AI agent architecture to match.
