- With its “central AI kitchen” now operational, Singtel has laid the initial groundwork and is now consolidating AI capabilities across divisions
- Singtel is measuring progress with three KPIs: financial impact, employee productivity and long-term initiatives
- The success of the company’s AI strategy will rely on addressing AI talent shortages, managing resistance to standardization, ensuring human-in-the-loop governance and driving large-scale workforce mindset change
Singapore’s largest service provider, Singtel, undertook a major restructuring exercise in mid-2025 to consolidate its AI initiatives – which had been spread across several departments – under one unit: AI and Data Analytics (AIDA). The initiative was launched to optimize AI investments, to ensure security and consistency as it scaled AI and to prevent duplicated efforts in different departments.
Kevin Yee, head of AIDA at Singtel, shared his vision for the project, some challenges his team is facing and insights about the initiative’s future.
Building the ‘central AI kitchen’
At the core of Singtel’s AI execution is what Yee calls the “central AI kitchen,” a common platform and framework designed to scale AI safely and efficiently across the organization.
The AIDA oversees this effort and has been operating for about six months now. During this time, it has completed initial groundwork. “We have finalized our organization structure in terms of people and their roles and responsibilities,” Yee said.
“The current focus is on consolidating existing AI capabilities developed across networks, IT and other business units; selecting the strongest components; and integrating them into a shared environment. Singtel is working with partners, including its subsidiaries NCS and RE:AI," he said.
The goal is to optimize resources and ensure consistency. “It is a centralized AI platform where we have common security guardrails that are put in place to develop and deploy AI at scale,” Yee said. Over time, this is expected to lead to a growing library of reusable AI agents that can be deployed across different parts of the business.
Measuring progress: ROI, ROE and ROF
Singtel is tracking the progress of its AI initiatives across three broad metrics: return on investment (ROI), return on employee (ROE) and return of the future (ROF).
ROI focuses on financial impact, including EBIT boost from new revenue streams and productivity gains, as well as cost reductions driven by platform consolidation and automation.
On the other hand, ROE measures how effectively AI tools are improving employee productivity and satisfaction. Singtel has rolled out Microsoft Copilot and, more recently, OpenAI’s ChatGPT Enterprise across the organization to boost AI adoption.
The third metric, ROF, captures investments in emerging areas where returns are harder to quantify now but could be strategically critical in the future. “For instance, autonomous networks where the return of investment is very hard to quantify,” Yee said. “Some of these areas are just too new. We invest in those areas even without proper ROI, in the hope that eventually some of these projects … will fall into ROI.”
Addressing talent crunch: Buy, build and borrow
Talent remains one of the biggest execution challenges. AIDA currently has around 60-65 people and plans to scale to roughly 130 within six months. The approach combines external hiring, internal reskilling and support from partners.
“We call it the buy, build and borrow,” Yee said. The company is actively recruiting while also upskilling employees from IT and network teams and augmenting capacity through hyperscaler partnerships.
Standardizing AI development across a large organization has not been frictionless. Yee acknowledged that resistance from business units has been part of the process. However, the fact that AIDA reports directly to the CEO has helped.
“There is some initial resistance. It’s not all smooth sailing. But having (the) CEO as our backer helps in smoothing all these conversations that we have with various teams,” Yee said.
Governance has become more critical as Singtel begins experimenting with AI-driven workflow changes. Allowing AI systems to take actions within live environments raises questions about accountability and risk. “If we let AI do everything, there may be serious consequences,” Yee said. As a result, Singtel believes that “human-in-the-loop is an important aspect of AI.”
What comes next
Several service providers – particularly in Southeast Asia, including SK Telecom in South Korea and Indosat Ooredoo Hutchison in Indonesia, among others – are trying to gain an edge by adopting AI in their operations.
Yee believes Singtel is slightly ahead in some areas, particularly data readiness. “The biggest problem telcos face in adopting AI is legacy systems and their data remains fragmented. All our data is consolidated in one place, and this gives us an edge over others,” he said.
“We are looking at ways to use AI to improve productivity across the organization, as well as to enhance the customer experience and how we serve our customers. But with SK Telecom, we are also exploring how to create LLMs customized for telco-specific use cases,” Yee said.
Looking ahead, Yee sees workforce transformation as the most difficult and most important challenge. “The technology … is just 30% of the whole AI story,” he said. “The (other) 70% is all about how we change people’s mindsets into adopting AI. The more difficult part is how to do workforce transformation so that they actually embrace and adopt AI in their daily work.”
As AI adoption deepens, Singtel expects workflows, roles and processes to evolve significantly.
Editor note: This story was updated on Friday, Jan. 9, 2026, at 9:26 a.m.