Tech Decoded : what Hyundai and Kia are really doing with artificial intelligence

• According to the group, H Chat Pro had more than 30,000 active users in July 2026. That represents about 80% of Hyundai Motor and Kia office and design employees, not the entire global workforce.

• An assistant helps engineers find old crash tests in the archives. Hyundai says that search work takes about 90% less time. It does not reduce the number of physical tests.

• The Atlas humanoid robots are not on the assembly line yet in 2026. They are still training. Hyundai wants them in factories as early as 2028, if the Korean union agrees.

A chat tool for employees

On August 12, 2026, in Seoul, Hyundai Motor Group presented its artificial intelligence update. H Chat Pro is the entry point for many employees.

Picture a very fast colleague who reads thousands of pages. You ask a question. The answer comes from company documents. The platform is connected to ChatGPT, Gemini, and Claude. Employees can also build small assistants there without writing code.

Tools that save time (according to Hyundai)

Hyundai’s figures mainly measure one thing. Finding information faster. Not replacing an engineer’s judgment.

Engineers crash cars in testing, then keep files, photos, and reports. An assistant helps find the right file faster. That is not 90% fewer real tests. It is 90% less time spent digging through the archives, according to Hyundai.

On the shop floor, driverless carts carry parts from one station to another, like small trucks. If one breaks down, the line can sit idle. Software then recalculates the routes. Hyundai cites about 86% less “unnecessary” downtime. In-house figure, not independently verified in public.

In the shop, a technician asks a question. A tool pulls up the useful pages from the manuals and the repair history. Hyundai claims about a 42% drop in response time.

A camera also automatically reads each car’s unique number on about 70 processes. Less typing by hand, fewer data-entry errors. About US$3.9 million in savings per year, according to Hyundai.

None of these percentages has public independent verification.

And the robots that look like humans?

Hyundai wants to go further. Factories more driven by software. And robots that walk like people.

The North American showcase is the Metaplant in Georgia. It assembles IONIQ 5 and IONIQ 9 models for the U.S. market, among other vehicles. Batteries too. Worth noting that in Canada, the IONIQ 5 and IONIQ 9 are assembled in Korea.

The through-line is Boston Dynamics. The American company is known for robots that walk and keep their balance, including the Spot robot dog and, now, Atlas. Since July 2026, Hyundai has owned 100% of it. The auto group also controls the company that builds these machines.

Atlas is a person-sized robot, with two arms and two legs. The idea is that it carries parts, puts them in the right order, then one day helps with assembly. In 2026, it is still training. It is not on the line. Hyundai is targeting about 30,000 Atlas robots a year by 2028. Announced timeline. Metaplant Georgia in 2028 for parts sequencing. Kia Georgia in 2029. More complex assembly around 2030.

Genesis Unveils the GV90, an Electric Flagship SUV with Available Coach Doors and a Four-Seater Configuration

The labour brake

In Korea, the Hyundai-affiliated union has, since January 2026, demanded an agreement before any humanoid robot goes on the shop floor. A machine that works day and night can replace several people. Hyundai says this plan is “not in” the current negotiations. The tension remains.

Reliability around people, costs, and social acceptance are still open questions. As long as those points stay unresolved, 25,000 Atlas robots remain a presentation target. Not a factory fact.

And for Canada?

There is no Hyundai or Kia assembly plant in Canada. No Atlas announced here. What can reach the customer first are technician-support tools.

The real question is not whether the internal chat tool gets adopted. It is whether Hyundai will turn those faster-search gains into a more automated factory, then into a fleet of humanoid robots. Under very real social and industrial constraints.