How APAC organisations are tapping generative AI

At Singapore’s Government Technology Agency (GovTech), a team of engineers is working on a chatbot powered by generative artificial intelligence (GenAI) to help citizens with career transitions and make better career choices.

Through a series of questions about a person’s interests and career journey, the chatbot would be able to help that person identify potential career opportunities, the skills required to make a particular career transition as well as available training courses.

Speaking to Singapore media on the sidelines of Google Cloud Next 2023 in San Francisco earlier this year, Chang Sau Sheong, deputy chief executive for product and engineering at GovTech, said the chatbot would help the government scale its career coaching programme which is being staffed by volunteers.

“We’re unable to scale [the programme] not only by the number of people, but also because career coaches don’t know everything,” he said. “This is one of our [GenAI] use cases, which is a little more unique than the run-of-the-mill chatbot as the intention is different.”

Another generative AI use case is in the booking of badminton courts which are in high demand across public sporting facilities in Singapore. Chang said Sport Singapore is looking to build a GenAI-powered chatbot to help people book badminton courts at certain time slots based on a set of parameters.

Over in Australia, employee engagement platform Culture Amp had been using its own AI models to analyse employee sentiment and comments for its customers, but it took its efforts further this year with large language models (LLMs) that summarise employee survey comments into topics and actionable insight.

The move is expected to help organisations automate a process that typically takes HR administrators up to hundreds of hours to complete. “The ability to bring the actual comments upfront and being able to synthesise the major themes very quickly means that HR administrators can take action much quicker than they would have otherwise been able to,” said Doug English, co-founder and chief technology officer of Culture Amp.

Across the Asia-Pacific (APAC) region, organisations like GovTech and Culture Amp have been doubling down on GenAI initiatives, more so than other parts of the world. According to a recent study by Enterprise Strategy Group and TechTarget, 75% of APAC respondents plan to adopt generative AI within the next 12 months, with nearly a third already running GenAI workloads in production or are testing the technology.

For every healthcare use case, NUHS will thoroughly test and validate generative AI responses, undertaken by multiple subject matter experts. We would also experiment with reasoning and cognitive features of generative AI, to further enhance its capabilities to support various use cases

The enthusiasm for generative AI in APAC is also reflected in IT budgets, with over half having allocated budgets to GenAI. Among them, 39% have allocated between 5% and 20% of their IT budget to the technology. The blinding speed of GenAI uptake among APAC organisations is also reflected in the 19% of organisations that are not yet sure if GenAI is a budget item.

Nevertheless, the rapid emergence of GenAI as a top IT priority is both impressive and alarming. The study shows that GenAI has become the fifth most important strategic initiative in APAC, trailing behind digital transformation, automation, cyber security, and cost-cutting, and surpassing traditional priorities like cloud and application modernisation.

That said, GenAI initiatives are expected to directly impact almost all strategic priorities in the near term, from automation and cyber security to cloud and application modernisation across various use cases. Currently, the top areas where generative AI is applied by APAC organisations include research, followed by marketing, IT operations, software development, and customer service.

In the future, marketing, customer service, and software development are expected to emerge as the top GenAI use cases for APAC organisations, driven by the desire to enhance conversational interactions with customers, reduce manual work for customer service teams, and expedite coding and code quality improvement.

Culture Amp’s English told Computer Weekly that it is currently exploring the use of GenAI in software development: “We haven’t adopted it heavily, but we do have some proof-of-concepts running. It definitely has potential but it’s still a very new area. We’ve been working with our security and legal teams to work through where and how we can leverage that technology”.

While Culture Amp is seeking legal counsel internally, APAC organisations overall are less likely to involve legal and compliance teams in their GenAI efforts than their counterparts elsewhere, according to the study. This could put organisations at risk, especially when they are still facing challenges with the lack of internal expertise, algorithmic transparency, and data quality, as cited by over a third of APAC respondents.

There are also concerns around intellectual property (IP) protection, according to Bhargs Srivathsan, a McKinsey partner and co-lead of the management consultancy’s cloud operations and optimisation work.

“There’s still a significant concern about IP infringement,” Srivathsan said, adding that while guardrails can be put in place for off-the-shelf models from cloud providers, there are other models out that are not fully tested for their ability to safeguard an organisation’s IP.

In APAC, 35% of respondents plan to leverage open-source LLMs and develop a GenAI solution in-house, higher than the global average of 30%, reflecting the region’s desire for more control over their data and IP. Additionally, 17% plan to develop a new LLM in-house, also higher than the global average of 9%.

Many enterprises think they need a Lamborghini to deliver a pizza. You probably don’t need a big and complex model with 65 billion parameters to generate scripts

Some APAC organisations are open to working with technology suppliers on their GenAI initiatives, with about a quarter indicating that they would work with a third-party provider offering access to a proprietary model or open-source LLMs that can be customised.

Singapore’s National University Health System (NUHS), for example, is working with Amazon Web Services (AWS) to test the use of Amazon Bedrock to build a GenAI solution to automate the creation of patient discharge summaries that document a patient’s stay in a healthcare facility, their diagnosis, treatment, and follow-up care instructions.

“Documentation is a particularly time-consuming effort for healthcare professionals, and automating this process will enable our clinicians to focus on their consultation with patients,” said Ngiam Kee Yuan, associate professor and group chief technology officer of NUHS.

“NUHS is at the forefront to develop and deploy generative AI technologies with precision and security built into the system. For every healthcare use case, NUHS will thoroughly test and validate generative AI responses, undertaken by multiple subject matter experts. We would also experiment with reasoning and cognitive features of generative AI, to further enhance its capabilities to support various use cases,” he added.

That NUHS is emphasising the need to test and validate AI responses mirrors the actions that APAC organisations are taking in AI governance. A quarter of respondents in the region are already developing methods to verify specific AI results for factual accuracy and conformity with business, ethical, and legal obligations, while over 40% require training on data security and privacy policies and procedures.

When it comes to costs, only 11% of APAC organisations indicated their willingness to pay over a 10% premium for a product or service that includes GenAI, compared with non-GenAI-equipped alternatives. Meanwhile, 43% of organisations were uncertain of their willingness to pay more, underscoring the importance of financial operations (FinOps) to manage the operating costs of GenAI workloads.

Srivathsan pointed out that many organisations that have experimented with GenAI were surprised by the associated costs and wanted more visibility into their workload expenses, with the choice of models being a critical cost factor.

“Many enterprises think they need a Lamborghini to deliver a pizza,” Srivathsan noted. “You probably don’t need a big and complex model with 65 billion parameters to generate scripts, for example. So, identify the right model that is cost-effective and efficient to generate what you really need. That can make or break your business case from the get-go.”