"In the short term, companies that create AI agents will win in the AI market. In the long term, companies that effectively combine existing services or products with AI are expected to dominate the market," advised Jang Jin-seok, a partner at BCG Digital, in a recent interview with Maeil Business Newspaper.
Partner Jang is a digital transformation expert leading BCG's digital and AI dedicated organization 'BCG X' at the BCG Seoul office. He has experience assisting clients with mid- to long-term digital transformation strategy consulting and actual business execution, including mobile application development. He has also conducted numerous productivity improvement programs utilizing AI. Recently, there has been controversy over whether AI-related businesses can secure profitability relative to high investment amounts. While AI semiconductor companies like NVIDIA are making money, companies that have launched AI services are still struggling to turn a profit. In this regard, Partner Jang stated, "AI service companies are still 'money-eating hippos' and are pouring enormous amounts of money into the situation," adding, "Companies like OpenAI, Google, and Perplexity have invested unimaginable funds into creating AI algorithms and large language models (LLMs)."
He explained, "These foundation model companies will eventually make money, but companies that directly provide AI services will be able to generate profits more quickly," noting that "the costs associated with using large models are rapidly decreasing, making it possible to secure profitability based on the reduced procurement costs."
Below is a Q&A with Partner Jang.
- How do you see AI companies making money?
▷ Currently, the way to make money is to create services that utilize AI and have customers pay for those services. However, in the future, the methods of making money with AI will become more sophisticated. There will be an increase in cases where profitability is secured by combining AI with existing services or products. For example, Adobe has integrated AI into existing services like Photoshop, allowing it to evolve from a company that simply sells solutions to a subscription-based service company that can retain customers. If this trend continues, IT service companies will eventually evolve into AI service companies, as IT companies will enhance existing profitability models with AI.
- In the short term, what AI services are expected to gain attention?
▷ Attention should be paid to AI agents. AI agents are automation tools that perform tasks based on generative AI. For instance, in the past, presentations were designed by humans. However, with the use of AI agents, initial presentations can be automatically generated. Reporters can then modify them based on this. In the short term, it will be important to see who creates which AI agent and gains actual customer preference. Especially in the case of business-to-consumer (B2C) AI agents, existing services will also gain customer preference through this.
- In which AI fields do Korean companies have relative strengths?
▷ The AI value chain consists of hardware, foundation models, domain-specific models, and agents. First, there is a question of whether Korea can create foundation models effectively. There are concerns about whether investments on the scale of hundreds of billions, like those made by OpenAI, Google, and Meta Platforms, are possible. Developing foundation models requires enormous computing power, which ultimately translates to costs, so if Korean companies recklessly jump in, it will be a significant challenge.
Ultimately, Korean companies should focus on services that are more closely connected to consumers within the AI value chain. As a manufacturing powerhouse, Korea should be able to create services like AI agents and integrate them into the manufacturing environment.
- What are your thoughts on the criticism that there is a lack of LLM models based on Korean data?
▷ While it is possible that better models could emerge from using only Korean data, the important factor is the data source. There are already concerns in the industry that open data used for training is being depleted. Consequently, there are calls to create customized LLMs for specific industries or domains. Even with current technology, AI's language proficiency is already established. Moving forward, it will be crucial to secure data for specific fields to train specific knowledge. In other words, the ability to create LLMs with expertise in domains rather than just language will be a key battleground.
In the past, AI was a single entity. However, it has already been divided into foundation models and fine-tuning models. This means that the technology to fine-tune the knowledge of language proficiency created by foundation models for specific purposes is important. This indicates that opportunities have actually increased, as specific domain-optimized AIs can also find opportunities.
- Are there any examples of companies effectively utilizing AI technology?
▷ There is the global pharmaceutical company Sanofi. Sanofi has a vision to become the first pharmaceutical company powered by AI. In fact, many efforts have already been made to shorten the drug development process using AI. Sanofi has this vision but has implemented AI more systematically. Sanofi has integrated AI into the entire pharmaceutical supply chain, including research and development (R&D) processes for finding new drug candidates, regulatory reporting processes, and clinical trials. Through this, Sanofi has achieved significant reductions in manpower and costs.
- What policies might President-elect Donald Trump, who has been re-elected, pursue regarding AI?
▷ Currently, while American companies are leading in AI technology, the importance of 'AI sovereignty' is increasingly coming to the forefront. This has led to a competitive landscape among countries regarding AI. Countries around the world, including Korea, China, India, and Europe, are contemplating ways to localize AI.
This presents a good opportunity for President-elect Trump. He values national interests, meaning he could frame the discussion around 'American AI.' There is a possibility that he will pursue quite AI-friendly policies, arguing that the U.S. must continue to lead in the AI field. Recently, discussions about 'Responsible AI' have emerged, but it cannot be ruled out that President-elect Trump may view this as a low priority and implement policies that allow U.S. AI technology to develop at a faster pace.
[Lee Jong-hwa, Reporter]
