The challenge

In pursuit of financial excellence and innovation, AIA’s Group Treasury Operations team embarked on a transformative journey to enhance its cash forecasting capabilities across the Group.

For a business of the scale and complexity of AIA, accurate and timely cash flow forecasting is a competitive advantage, especially in dynamic financial markets. However, traditional methods, often reliant on manual data analysis and static models, have inherent limitations. This can lead to suboptimal cash flow forecasts, resulting in conservative cash positions which limit a business’s ability to capitalise on income-generating opportunities fully.

Through AIA’s Technology, Digital and Analytics (TDA) strategy, the insurer pioneered artificial intelligence (AI) technologies to address common forecasting challenges and take its treasury capabilities to next level.

The solution

AIA Group Treasury Operations developed a proprietary predictive cash forecasting model which has been successfully piloted within the businesses. The model applies AI machine learning algorithms to forecast major cash line items for operating entities.

Data from AIA’s administrative systems is first funnelled into a centralised hub, the Treasury Data Mart, where it is curated and fed into the AI model. This produces one-month forecasts which are integrated into the daily cash forecast processes of the insurance businesses. The treasury team can then overlay expected flows to enhance forecasting accuracy further.

Historical ex-post actual cash balances are entered into the AI model, enabling it to adapt and refine its predictive capabilities over time. This feedback loop enhances the model’s forecasting precision.

“Harnessing the combination of machine learning and domain expertise through manual operator intervention is extremely powerful. Our group treasury has observed up to 90% more accurate forecasts on specific cash flow items,” says Daniel Christie, Head of Treasury Operations.

So far, the pilot has been deployed in the following AIA markets:

• Malaysia (in Phase 2 production).

• Philippines (in Phase 2 production).

• Singapore (model trained in Phase 2).

• Indonesia (model trained in Phase 2).

Best practice and innovation

The retrospective action of feeding the model actual data allows it to refine its forecasting accuracy continuously. This best practice is important to remain ahead of market dynamics. In addition, the predictive machine learning relies on ex-post observations to overlay ex-ante inputs from the treasury team who have a line of sight on near-term unexpected cash flows. This creates a powerful and innovative solution by combining both humans and machines.

Key benefits

• Advanced modelling and analytics can identify opportunities for optimising cash balances and generating additional investment income.

• Smart automation can reduce manual operation (labour cost savings and operational risk mitigation).

• Machine learning models can reduce the time to produce forecasts (process efficiency).

• Intelligent algorithms improve forecasting (accuracy improvements).

• Platform modernisation (Power BI and Power Apps) enhances the visibility of liquidity across the Group.

• Data mart solutions provide durability and economies of scale.

• Data governance regulates processes and controls for data quality.

“We are very proud to have developed this solution in-house. Our cash flow forecasting is more accurate and better optimised and we are just beginning to unlock the potential benefits of AI to drive our predictive cash flow forecasting model.”

Daniel Christie, Head of Treasury Operations