[Economic Review Interview]

"AI is only valuable while it's in your hands. Once it leaves your hands without reaching customers, it becomes technology that no one needs. Our job is to put that technology into the hands of the people who can actually use it."

In an interview with Economic Review on June 22, A.I.MATICS CEO Hoon Lee shared his vision for building AI that creates real business value—not by competing head-on with Big Tech's foundation models, but by delivering AI solutions that solve immediate problems in real industrial environments.

His ambition is clear: transform A.I.MATICS into a Physical AI company capable of supporting work traditionally performed by factory operators, healthcare professionals, and many other frontline workers.

From Korea's First Corporate Venture to Vision AI

A.I.MATICS traces its origins to Korea's first in-house startup launched within Hyundai Motor Group after three KAIST graduates developed pioneering automotive technologies.

"Our very first product was the world's first lane departure prevention system," Lee recalled. "Today everyone talks about deep learning, but back then our engineers literally drew lane markings by hand to make the technology work."

For more than fifteen years, the company supplied lane departure warning systems and ADAS technologies. Then everything changed.

Around 2017–2018, major OEM contracts shifted to overseas competitors, forcing the company to rethink its future.

Rather than continuing as a component supplier, A.I.MATICS decided to build its own AI platform. As deep learning rapidly emerged, the company invested heavily in Vision AI. When Lee joined four years ago, the team focused on applying those capabilities to mobility, allowing them to enter the market much more quickly.

One insight became particularly important: driver behavior is directly connected to accident rates.

"Traffic accidents ultimately lead to higher insurance costs," Lee explained. "We asked ourselves, what if AI could go beyond detection and become a tool for driver coaching and fleet management?"

That idea became the foundation of the company's AI Safe Driving Service.

The system analyzes behaviors such as distracted driving, traffic signal violations, and blind-spot events detected on edge devices. After each trip, drivers receive a safety score, while fleet managers receive reports that help guide coaching and training.

The hypothesis proved correct.

After launching with Sampyo Ready-Mix trucks two years ago, the solution has since been deployed across approximately 3,000 Samsung commuter buses, along with fleets operated by Hyundai, LG, and Samsung Logistics.

"Within Samsung's commuter bus fleet, driver scoring reduced accident rates by roughly 40–50% year-over-year while lowering insurance premiums by around 20%," Lee said.

The platform has also expanded into regional governments and intercity bus operators, including the AI safety platform now deployed across Gyeonggi Province's public bus system.

The solution has gained particular traction in logistics, where reducing accidents directly lowers cargo damage and insurance costs.

A Powerful AI Model That Fits Inside a $20 Chip

At the heart of A.I.MATICS lies its proprietary AI engine, aimNet.

Rather than relying on massive foundation models or shared AI architectures, aimNet is designed to build lightweight, efficient, and highly specialized models optimized for specific industrial applications.

"General-purpose AI models can do many things, but they're rarely economical for vertical AI," Lee explained.

"A large model often requires GPU hardware costing several thousand dollars. We instead deploy our models inside NPUs that cost roughly twenty dollars, coding them specifically for the customer's environment."

Instead of recognizing everything, aimNet focuses only on the objects and situations customers actually need.

The technology evolved from the company's original Roadscope platform, combining years of computer vision expertise with deep learning.

Roadscope has now reached its tenth generation.

"Deep learning was introduced beginning with Roadscope 7 and further refined through Roadscope 8 and 9. Those technologies eventually evolved into today's AI Safe Driving Service."

Roadscope 10 represents what Lee calls the complete vision of AI-assisted safe driving.

"A bus driver must constantly monitor passengers, blind spots, and surrounding traffic simultaneously," he explained.

"Each camera requires its own AI engine. Roadscope 10 DVR essentially becomes an additional set of AI-powered eyes for the driver."

Unlike conventional systems, Roadscope 10 not only issues real-time edge alerts but also uploads every event to the cloud, allowing fleet managers to review, evaluate, and coach drivers after each trip.

Asked about the company's competitive advantage, Lee answered confidently:

"Our technology is small, powerful, and economical. What more could customers ask for?"

Still, he made clear that A.I.MATICS has no intention of competing directly against foundation models.

"Our goal isn't to challenge Big Tech. Our goal is to commercialize AI that customers actually need in the field."

One Core Technology, Multiple Industries

One of A.I.MATICS' defining strengths is that a single Vision AI technology powers multiple industries.

The same core AI that enables safe driving on the road is also used for automated visual inspection in manufacturing.

Lee sees manufacturing as the company's largest long-term opportunity.

"After PCB assembly, there are still many inspection steps performed manually," he explained.

"Our AI replaces human visual inspection by automatically identifying assembly defects."

In other words, the same "eyes" that monitor moving vehicles now inspect stationary production lines.

The company is also expanding into healthcare through ECG patch solutions being developed with parent company Dreamtech, while simultaneously developing chatbot and voicebot solutions powered by its proprietary Small Language Model (SLM).

The SLM is being developed together with Insung Data, whose platform supports roughly 60% of Korea's quick logistics market.

"Our voicebot solutions are designed to automate call center operations," Lee said.

Turning AI Research into Revenue

Despite operating across multiple industries, Lee rejects the idea that the company's strategy lacks focus.

"Very few companies actually make money from core AI technology alone," he said.

"Our priority is expanding the technologies we already excel at and applying them directly to real-world businesses."

Financial results support that strategy.

The AI Safe Driving business is expected to grow from approximately KRW 4 billion in revenue last year to roughly KRW 12 billion this year.

Vision inspection is projected to increase from KRW 2 billion to approximately KRW 6–7 billion.

Healthcare development projects have already generated cumulative development revenue of around KRW 2 billion.

Meanwhile, the chatbot business follows a usage-based pricing model, while voicebots operate under monthly subscription plans—allowing new products to generate revenue quickly after commercialization.

"Our biggest strength is that our technologies can become revenue almost immediately," Lee said.

Building AI Globally

The company's rapid expansion has also been supported by strong group synergies.

After becoming a subsidiary of Dreamtech in October 2024, A.I.MATICS has accelerated growth by leveraging the group's manufacturing ecosystem.

Most recently, the company established its Vietnam subsidiary to support growing demand for AI-powered manufacturing inspection.

"Dreamtech and Namuga are Tier-1 suppliers to Samsung, so much of their production already takes place in Vietnam," Lee explained.

"We expect to expand from visual inspection into manufacturing facilities across Bac Ninh and Hanoi."

Healthcare expansion follows a similar strategy.

Dreamtech's U.S. subsidiary, Cardiac Insight, provides global distribution channels through which A.I.MATICS plans to integrate its deep learning technologies into ECG patch solutions.

Cutting AI Deployment from Three Months to Two Weeks

Another key competitive advantage lies in A.I.MATICS' remote AI deployment framework.

Lee recalled early lessons from entering North America.

Traffic rules and road environments differed significantly from Korea, making locally trained AI models perform poorly in the field.

Initially, completing new datasets required nearly three months.

The solution came through AWS-based remote deployment.

"We remotely push new AI models to the field, collect new data through the cloud, retrain our models, and redeploy them remotely," Lee explained.

"What once required three months can now be completed in roughly two weeks."

The same workflow now supports factory inspection systems.

Rather than traveling to overseas production sites, engineers collect manufacturing data remotely, retrain models in Korea, and immediately distribute updated AI models through the cloud.

Today approximately 9,000 commercial vehicles across Korea use the company's AI Safe Driving platform.

A $100 Billion Smart Factory Opportunity

Although mobility remains an important business, Lee believes manufacturing will become the company's primary growth engine.

"The mobility market is highly competitive," he said.

"We deliberately position ourselves in the B2B aftermarket."

"Meanwhile, the smart factory market exceeds KRW 100 trillion, and visual inspection represents one of the final stages of dark factory automation."

The company also plans to combine AI-powered visual inspection with Autonomous Mobile Robots (AMRs), creating an integrated automation package for global manufacturers.

Rather than competing on hardware, A.I.MATICS focuses exclusively on AI software and vision technologies while partnering with hardware manufacturers.

"Competing against Chinese hardware companies would be a losing battle," Lee said.

"Our competitiveness comes from Vision AI and software."

Toward Physical AI

Ultimately, Lee believes A.I.MATICS is evolving into a Physical AI company.

"Vision AI naturally leads to Physical AI," he said.

"The first step is what we're already doing today with AMRs."

He sees years of accumulated Vision AI expertise as the essential foundation.

"Human vision cannot be replicated overnight. It requires years of data and experience."

That foundation, he believes, will eventually extend beyond factories.

"In hospitals, Physical AI could attach medical patches, deliver sterilized equipment, or perform many routine nursing tasks," Lee explained.

Whether on roads, factory floors, or inside hospitals, Lee envisions the same core Vision AI technology becoming the eyes of intelligent machines.

His vision is simple:

Build AI models that are small, powerful, economical—and capable of transforming how work gets done in the real world.

Source : Enonomic Review ((https://www.econovill.com)