By now, nearly every organisation has heard of AI. While everyone is eager to leverage its potential, few have laid the necessary groundwork to fully realise its benefits.
Crafting a solid AI strategy is much like building a house — it requires a strong foundation, careful planning, and thoughtful execution. Without these, the AI journey can falter before it even begins.
This sentiment was echoed during the “Data Modernisation: Transforming into an AI-Powered Organisation” panel at Databricks’ Data + AI World Tour Singapore event, where senior data and AI experts shared the challenges they faced and strategies they employed to drive AI initiatives within their organisations.
Data culture
How exactly can technology make a tangible difference in today’s enterprises? For Andy Ta, Director, Data Analytics & AI and Chief Data Officer, Synapxe, it all begins with having a modern data platform that users can easily access to simplify their workload.
“For healthcare, it took us many years to secure the data, but we also need to tap into something that allows for experimentation. Additionally, we should be able to share the data with collaborators and partners so that it can be used at a larger scale to create more value,” Ta said.
Ritchie Ng, Research Scholar at NUS and Head of Data & AI at Eastspring Investments, pointed out that many organisations lack data governance, which is causing AI deployments to fail.
“It’s not just that their data is fragmented across different systems; their processes are also inconsistent and unoptimised across various functions and countries, which presents a significant problem,” Ng observed.
Drawing from his experience, Ng outlined three steps they took to address their data challenges: “First, we optimised all our processes, then we standardised them. After that, we applied a unified data and AI platform. This gave us a platform where you could deploy any algorithm for a single use case and expand it both horizontally and vertically.”
For the Central Provident Fund (CPF) Board, systems and processes are only as effective as the people managing them.
“You may have the perfect system, but it can overwhelm the people using it. For us, it’s about educating our end users and upskilling them. Databricks has done a great job lowering the barriers. We conduct our own courses to train users, and before they onboard the platform, we ensure they understand the importance of keeping financial data secure,” shared Vance Ng, Director of Data Science Accelerator at the CPF Board.
For Bharathi Viswanathan, Chief Digital & Information Officer at Suntory, technology should be adopted to address a business need, not simply for its own sake.
“You have to look at your employee in the context of their entire role and really focus on what superpower you are uncovering for them. The question you want to answer is, ‘What superpower am I giving this employee through technology?’” she said.
ROI challenges
To get the Chief Financial Officer to approve any AI budget, there must be a clear return on investment (ROI) within a specified time frame. However, according to Synapxe’s Andy Ta, ROI isn’t just short-term financial gains.
For example, he shared how he secured a small budget to enable generative AI access for 90,000 healthcare workers. The aim was to foster experimentation and innovation for potential healthcare use cases.
“We’re getting users to create prompts and experiment, and if the use case works for me and a few others, we can then productionise it,” he said.
From Ritchie Ng’s perspective, ROI on process optimisation and standardisation may not be immediately obvious to enterprises, but it pays off over time.
“You will see immediate cost gains the moment you expand and head straight for markets across the region,” he noted.
For government entities like the CPF Board, ROI can be harder to quantify. Yet, as Vance Ng explained, their mission and their policy outcomes are clear — to secure retirement adequacy for all Singaporeans.
“It’s difficult to set individual ROI targets for each proof of concept, but in terms of policy delivery, our ROI is clear. When we implement a policy, we design it based on the data we have. Then we’ll know if the client is satisfied — the immediate feedback is our ROI. In terms of operational productivity, measuring ROI can be challenging, but we look at how processes have improved. We assess the time taken previously and, with revised processes, aim to optimise it,” he said.
Improving services
At the core, every organisation aims to improve service delivery for their stakeholders — whether it’s a patient visiting a hospital for a consultation or someone accessing government services. AI can enhance programs and processes to provide greater speed, efficiency, and convenience, but it is not a panacea. Most importantly, these advancements must not compromise personal data.
We need to be efficient and focused. We must determine the best way to serve our members and uphold trust in the government, not just in the CPF Board. When educating our data stewards and data owners, who are typically heads of departments, we ask them questions like ‘What kind of use cases do we need to account for?’” Vance Ng stressed.
“For me, I tend to focus a lot on people. Processes and platforms tend to take care of themselves,” he concluded.
