IBM launches Gen AI for Thai enterprises

Nophakhun Limsamarnphun

Thailand’s international competitiveness will jump if a significant number of its enterprises adopt artificial intelligence (AI) tools to boost productivity, according to Anothai Wettayakorn, managing director of IBM Thailand.

At present, it is estimated that only 5-6 per cent of Thailand-based enterprises have turned to Generative AI (Gen AI) solutions to increase their competitiveness, compared to the global estimate of 10 per cent. Over time, if Thai enterprises could achieve an adoption rate of 15-20 per cent, the country will become much more competitive internationally.

Anothai said Gen AI is not a hype but a core technology for enterprises, ranging from banks and insurance firms, telecom, retail to healthcare, education and manufacturing, among other sectors. In Thailand, banks, telco, and retail firms have already piloted Gen AI projects for predictive, customer service, coding, machinery maintenance and other tools.

Speaking at a media roundtable on the next frontier of AI race in Thailand on Nov 8, he asserted that AI and automation investment should not be seen as a cost but tools to do more with less, especially for enterprises to meet new business and economic challeges such as a slower economic growth and increased competition.

Even small enterprises could consider investing in AI tools which now start at only about Bt2 million, he said, adding that enterprises should start planning AI adoption by formulating a good AI strategy and use cases to increase their productivity and competitiveness.

There are now a variety of Gen AI tools such as geospatial tools suitable for agri tech, time series for banking and finance, predictive tools for retail such as The Mall group while Bank of Ayudhaya (Krungsri)  and  TMBThanachart Bank have adopted Gen AI tools for coding; Provincial Electricity Authority, for customer service and IRPC for machinery maintenance.

Ecosystem integration is a key factor for successful implementation as only 10% of AI pilot projects are deployed worldwide, whereas 30% of pilots are dropped off. Thailand is currently consistent with the global trend, with people, budget, regulations, tech readiness and safety & security also among key success factors.

Many CEOs are still reluctant to launch AI projects largely due to concerns about potential risks. However, the interest to familiarize themselves with Gen AI for enterprises was huge as over 1,300 registered to attend the latest IBM event in Bangkok in early November.

Anothai said retail enterprises, for example,  can use Gen AI to plan their year-end promotion campaigns more effectively or provide more personalized shopping services to target customers when executives are equipped with Gen AI tools for decision making --which was previously not practical.

In other words, management can make more precise, effective and timely decisions with the help of Gen AI models and massive datasets to boost productivity for more customer satisfaction at a lower operating cost using a smaller-fit-to specific-purpose AI model than those used by general purpose models for consumers such as ChatGPT which relies on a Large Language Model (LLM) of 80-100 billion paarameters compared to only 2-8 billion parameters for enterprise Gen AI models.

For example, a data set and model of 2-8 billion parameters is practical to drastically improve the automotive insurance claim service as AI tools can be deployed like augmented assistants and agents using digital labour rather than manual work. Smaller-to-fit models are much cheaper to deploy and operate compared to consumer-oriented models using huge datasets, massive models, massive computing power and other resources.

Another example is a Gen AI model for bank loan approvals which can rely on a smaller-to-fit-specific-purpose model using in-house datasets,  public data set from the credit bureau and other data to deliver a higher quality service for customers.

For enterprises, IBM has launched its Granite 3.0 for smaller-to-fit-purpose model of 2-8 bn parameters, bringing the massive investment and operating costs of LLM down to a manageable level for large, medium and small enterprises --due to less use of the more expensive GPUs when compared to CPUs.

Anothai said both CEOs and CFOs have to collaborate when considering the return on investment (ROI) on AI pilots, while enterprises should better aim for drastic improvement in productivity, rather than only incremental improvement.