CBA’s AI solution
Recognising that customers value having a complete picture of when their payments are due, CBA developed Bill Sense. This AI solution provides CommBank app users with greater insight into their saving patterns and bill frequency. Customers who use Bill Sense see how much money they’ll need to cover their bills each month.
Bill Sense uses advanced machine learning to:
- understand the patterns of customers’ regular payments
- predict when future bills will arrive and how much they will be.
This is achieved by learning from customers’ previous transactions. No information is shared or obtained from billers. Customers have total control of the parameters and information that Bill Sense uses to make its predictions. This includes:
- whether to include a specific biller
- setting the payment frequencies
- adjusting the predicted payment amount.
Considering AI ethics
Strong data and risk governance processes support the design of Bill Sense to ensure it operates safely and with appropriate accountabilities.
These ensure that the product:
- benefits customers
- aligns with the bank’s established ethical framework, which is compatible with the Australian AI Ethics Principles.
Putting AI ethics principles into practice
CBA shared the following information about how Bill Sense considers the Australian AI Ethics Principles.
Principle 1: Human, social and environmental wellbeing
CBA’s purpose is to improve the financial wellbeing of its customers and communities. The team designed Bill Sense to help customers meet their financial commitments and manage any stress around bills.
Principle 2: Human-centred values
Bill Sense is a purposeful digital solution that can help to solve genuine customer problems. Bill Sense can help alleviate any feelings of helplessness for customers who may need assistance managing their bills.
Customers control the inputs that inform the AI predictions. They direct when and how they use Bill Sense. When they choose to opt in to use it, they can override or remove any predicted bills.
Principle 3: Fairness
Bill Sense is accessible to all users and operates the same way regardless of individual personal or financial characteristics. The CommBank app uses information the same way for all users, which reduces the likelihood of bias or unfair discrimination. Customers can provide negative or positive feedback on Bill Sense directly through the app.
CBA‘s policies and guidelines require machine learning models to not use any variables that may lead to unintended discrimination. Bill Sense uses only the minimum information it needs to operate.
CBA also has an Ethical AI service to investigate their models for any disparities (for example, to check a model is not behaving undesirably for different groups of customers). The service is relevant to the principle of fairness, as well as transparency and explainability. CBA modellers and AI practitioners use it to extract explanations of outcomes from complex AI models. This lets them:
- evaluate disparities in decisions
- obtain explanations of what variables are normally important to make predictions and what variables are used for each individual prediction.
Modellers can use this to sense-check their model and its predictions.
Principle 4: Privacy protection and security
Strong data protection policies ensure that data use is approved and only used for the purpose it was originally collected.
Bill Sense only learns from customers’ previous transactions within their account. No information is shared or obtained from billers (for example about a customer’s bill or account) to predict future bills.
Principle 5: Reliability and safety
Strict governance practices ensure that data sources, systems and models are reliable and safe.
For example, staff build all AI systems in accordance with the CBA’s Model Risk Policy. This policy ensures teams register and periodically review models to check they are operating in line with their intended purpose.
The AI system generates alerts allowing humans to check and rectify any problems that may occur. This also enables the system to be taken offline for the team to resolve issues or revert to previous versions if needed.
Principle 6: Transparency and explainability
The CommBank app, which houses Bill Sense, informs customers that it predicts bills based on their previous transactions. This means predictions of future bills may not always be correct.
Customers are informed of this risk and are always in control to use, override or disregard a predicted bill.
Principle 7: Contestability
Bill Sense is part of the CommBank app, so customers can provide feedback within the app. The app directs any complaints to responsible staff, who will take action when needed.
Principle 8: Accountability
CBA’s governance model ensures there are accountable parties for all systems, data points and the AI Model.
CBA clearly identifies accountabilities in the governance of every system and its components. These accountabilities include:
- business owners
- service owners
- any other responsible team.
Benefits and impacts
CBA had an ethical framework for decision making in place when it joined the AI Ethics Principles pilot. This framework aligns closely with the majority of the Australian AI Ethics Principles.
CBA Chief Decision Scientist, Dan Jermyn, said:
‘We see the government’s principles as important as they make explicit things like fairness, transparency and explainability which can be overlooked when developing complex AI systems.
‘Our organisation is committed to using technology, data and AI safely and with the right accountabilities to improve our customers’ financial wellbeing.’
If you have any questions about Bill Sense or or CBA’s AI systems or ethical AI journey, please contact email@example.com