Cohere, a Toronto-based enterprise AI company, is aggressively expanding its business across the Asia-Pacific region. In a recent interview with IT Chosun, Jang Hwa-jin, Cohere’s APAC managing director, emphasized the company’s on-premises deployment strategy, stating, “The core of sovereign AI is whether customers can maintain control over the model and use their data securely.”

Cohere’s definition of sovereign AI places emphasis on practical control rather than the nationality of the model. “Once Cohere’s AI model is installed at a customer site, even Cohere cannot know what data is being processed,” Jang explained. “There is also no kill switch that could suddenly disable AI usage from the outside.”

Founded in 2019, Cohere was co-founded by Aidan Gomez, a co-author of the paper that introduced the transformer architecture—the foundational technology behind generative AI. The company has focused on supplying AI models and agent platforms for enterprise and government use rather than consumer chatbots. In July 2025, Cohere opened its APAC hub office in Seoul and recruited Jang to lead the region. In April of this year, the company also partnered with German AI firm Aleph Alpha to target the European market.

Targeting Regulated Industries…The Security Advantage of On-Premises

Cohere is targeting regulated industries where data cannot easily be transferred externally, including the public sector, finance, semiconductors, aviation, and shipbuilding. Its AI models and the agent platform “North” are installed on customers’ GPU servers or isolated virtual private clouds (VPCs), and can operate in air-gapped environments with no internet connection. Since data is processed only within customer-designated infrastructure, Cohere itself cannot access customer data.

Jang pointed to the U.S. Cloud Act as a key driver of growing on-premises AI demand in regulated industries. Under the Cloud Act, cloud services operated by U.S. companies must comply with U.S. government requests for information, meaning companies holding nationally critical technologies feel burdened simply by having their data processed by foreign entities or connected to the cloud.

Results in the APAC region are becoming visible. In South Korea, Cohere completed a project with Hanwha Ocean. In Japan, the company won projects with the Ministry of Justice and Ministry of Defense alongside Fujitsu. It recently supplied North to Mitsubishi Heavy Industries on a pilot basis, with a full contract expected soon. In Singapore, Cohere signed its first contract with a Ministry of Defence-affiliated agency in March and secured an additional contract within three months.

Flat-Fee Pricing Provides Cost Predictability

Cohere’s on-premises deployment contracts are billed on an annual flat-fee basis rather than by token usage. The flat annual fee covers not only LLMs but also updates to “Embed” for document search and “Rerank” for search result re-ranking, as well as North feature updates and technical support. Infrastructure costs such as GPU servers are borne separately by the customer.

Jang emphasized that this structure is attractive to finance executives. “There was a recent case at Uber where a planned year’s worth of AI budget was exhausted within a few months,” he said. “With on-premises deployment, costs are predictable, so employees can use AI 24/7 without concern.” He added that some North American customers have chosen Cohere specifically because token-based pricing became burdensome.

North can be integrated via API into internal portals or employee-facing solutions, and agents can be configured with just a few clicks—no coding required. The next version will add the ability to build workflow automations using natural-language voice commands alone. North is also embedded as a studio module within LG CNS’s agent platform “AgenticWorks.”

South Korean Entity to Be Established This Year…Targeting Triple-Digit Customer Count

Cohere has been ramping up its South Korean business since 2025, built on its partnership with LG CNS. A representative example is LG CNS fine-tuning Cohere’s base models for Korean and jointly participating in a Ministry of Foreign Affairs project. “Recently, we’ve also been signing multiple private-sector contracts,” Jang said. “I can’t disclose names yet, but Cohere is independently developing business and seeing results.”

The number of new APAC customers has already grown fivefold this year compared to last year, with plans to double that again by year-end. The goal for next year is to double once more, bringing the customer count to triple digits.

Headcount is also being expanded aggressively. Cohere plans to double its workforce by year-end compared to June, focusing on hiring talent proficient in both English and local APAC languages. The company is placing particular emphasis on recruiting IT technical staff, including field deployment engineers (FDEs) and solution architects.

The establishment of the South Korean legal entity is in its final stages. “The incorporation process itself can be completed within a few weeks once documents are submitted, but preparations required for a foreign company—such as opening bank accounts—are taking time,” Jang said. “We will complete the South Korean entity this year and then proceed with establishing a Japanese entity.” Once the entity is established, the inconvenience of having contracts handled directly by the U.S. or Toronto headquarters will be resolved, making direct contracts with customers in South Korea and other countries significantly easier.

Jang also outlined plans to expand the domestic partner ecosystem. “LG CNS is a core partner, but we need to expand specialized partners to serve customers in different industries going forward,” he said. As APAC head, he envisions expanding into markets including South Korea, Japan, and Singapore, as well as Australia, India, and Taiwan. Cohere recently signed a contract with a local reseller in Taiwan, and plans to begin hiring in India and Australia by year-end or early next year to accelerate business expansion.

Meanwhile, LG CNS is supporting customers in leveraging Cohere’s collaborative models through AgenticWorks, tailored to their environments. AgenticWorks is a platform that supports the full lifecycle of AI agents—planning, design, build, deployment, operations, and management—comprising six modules: Refiner, Knowledge Lake, Router, Builder, Hub, and Studio. LG CNS is also conducting technical validation with FuriosaAI to apply the South Korean-made NPU “RNGD” to the AgenticWorks runtime environment.