Organisations will have to figure out what kind of workforce they want to manage as artificial intelligence (AI) agents take more prominence in their future operations.
This includes having to answer questions about how much they want their AI agents to take over, according to Yap E. Fang, chief architect and practice lead for cloud, data, and AI for Asia-Pacific, at NCS. The Singapore-based IT services vendor is a wholly-owned subsidiary of Singtel Group.
Fang noted that enterprise customers increasingly have brought up questions about how agentic AI will impact their workforce.
“With agentic, what has been interesting is the conversation around the extent [organisations] want to replace [their] human workforce with AI agents,” she said in a video interview with FutureCIO. These have encompassed questions around whether it is time to shut down IT departments or what is the new role of IT departments, she noted.
"And whether it’s time to make such decisions,” she added.
These are valid questions, Fang said, as companies will have to think about how they want to manage an AI agent workforce.
“For me, I would look at the job to be done and who is fulfilling it,” she said. “Where is the accountability? Can the AI agent be fully accountable? If not, how then should the job or role be redefined, such that the final [onus] is on the human.”
Getting humans to work with AI agents
Organisations then should set the right expectations amongst their employees and provide the necessary training to facilitate a workforce that is empowered by AI agents, Fang said.
For example, new skillsets may be required to help human employees manage AI agents and fine-tune them to perform better.
Getting an AI agent to be the manager, though, remains a novel concept that has yet to be adopted, Fang said.
In fact, 83% of workers in Singapore said they would be comfortable working alongside AI agents, but just 8% said they were comfortable being managed by an AI agent, according to a recent survey by Workday.
Globally, 75% said they would be comfortable to team up with AI agents, but 30% were comfortable being managed by one, the study found. The Workday report polled 2,950 decision-makers and software implementation leaders in Asia-Pacific, EMEA, and North America.
In Singapore, 79% said their organisation already were rolling out AI agents or had begun operating them.
Some 24% of respondents in the nation-state said they were comfortable with AI agents operating in the background without human knowledge.
Another 90% believed agents would help them get more done, though, 50% expressed concerns the productivity gain would drive down critical thinking. Another 33% pointed to concerns about a drop in the quality of human interaction.
All respondents believed some form of regulation was required for AI agents, with 57% supporting ethical guidelines established by developers. Some 48% highlighted the need for strict human oversight.
“We’re entering a new era of work in Singapore where AI is an incredible partner to organisations today, complementing human judgement, leadership, and empathy,” Jess O’Reilly, Workday’s Asean general manager, said in the report. “To drive productivity and trust, it is important that we rely on AI as a partner rather than a leader.”
“To do so, Singaporean organisations need to be intentional in how AI is used and keep people at the centre of every decision,” O’Reilly said.
Go hands-on to keep skills sharp
This also may mean getting them to skip the AI sometimes and choosing to still carry out tasks manually.
Pointing to the need to stem any potential decline in human skills, Fang said she encourages her team of software developers to turn off any AI-assisted coding tools, when there are no looming delivery deadlines, and take the time to think about how they want to design a piece of code.
“This is important as it forces them to have the discipline to think and not just [depend on AI-powered] simulations,” she said.
Doing this also helps mitigate concerns about any progressive loss of skills as AI use increases.
Asked about her thoughts on vibe coding, Fang said she pushed her team to trial the concept and determined the technology was not ready.
First coined in February 2025 by OpenAI cofounder Andrej Karpathy, vibe coding describes a completely hands-off approach to software development, powered entirely by AI.
The model may be useful for teams that have to churn out multiple prototypes and proof-of-concepts, said Fang.
Vibe coding can help accelerate the production of such projects and get things started, for example, pushing out different UX (user interface) designs so companies can assess the possibilities.
It can cut down the production time from weeks to overnight, she said, adding that this can be useful especially for projects that have to generate mixed media, such as graphics.
However, the technology that powers vibe coding still is not sufficiently advanced or robust for more complex algorithms -- at least, for now, she noted.
Her team of software architects currently use AI-assisted tools to aid in their coding, including NCS’ Sunshine.Coder, the vendor’s AI-assisted software development platform. The AI offering is touted to help developers review and test codes.
“The key thing I emphasis is, whatever technology we pursue, we need to know the value it brings, and with that, the problems we’re trying to solve,” Fang said.
She brings a similar message to NCS’ clients, including those that are eager to adopt every new version of a popular AI tool as soon as it is released.
Organisations should keep in mind the ultimate outcome they want from implementing such new technology, so they are more conscious about the reasons they want to use the tool, she added.
It also means they need to measure the results, which often is overlooked, she noted.
Companies need to know if the solution they implemented is working or risk having their employees slip back to habits and reverting to the old ways of working, she said.
