How to Build Trust When Using Generative AI & Chatbots To Improve Digital Financial Services?

Imagine this: just like every month, you’ve just paid your rent through your digital wallet. But then, your landlord calls: They haven’t received a thing. The money’s gone from your account, yet it’s nowhere to be found. So, what now? Who do you turn to for answers and action?

If you lived in the Philippines, you could write to BOB. BOB is the primary channel of the Bangko Sentral NG Pilipinas (BSP) Consumer Assistance Mechanism. It is powered by AI and natural language processing technology, allowing it to handle English, Filipino, and Taglish complaints. BOB is available 24/7, accessible via webchat on the BSP website, SMS, and via Messenger on the BSP Facebook page.

According to data from the BSP, 43,115 complaints were filed and processed in 2023, and around 87.4 per cent of those were processed through BOB. One research study found that the chatbot can categorize 74 per cent of the complaints, saving substantial time for BSP staff members, who previously had to parse through the complaint data to classify each complaint manually.

BOB is an example of how conversational AI can effectively enhance efficiency and customer experience in the digital finance sector. AI chatbots could also provide financial counseling to underserved populations. However, the rapid rise in popularity of these AI tools raises essential questions about how players in the digital finance sector should implement this type of technology responsibly and without hurting users’ trust.

A Rising Trend Across the Digital Financial Services Industry

The public release of ChatGPT in November 2022, now considered the fastest-growing consumer application in history, has created a wave of adoption of conversational chatbots that promise to deliver more comprehensive interactions by providing multilingual support, sentiment analysis, personalized responses, and direct connection to companies’ knowledge bases.

Firms like OpenAI offer a direct connection to their models, and the banking industry has already rushed in. For example, the bank Santander in Mexico has been using ChatGPT for customer service since 2023, but only to provide essential customer orientation and not to reply to account-specific requests.

In Colombia, Davivienda is leveraging Generative AI to enhance customer service through its DaviPlata platform. By integrating voice recognition and natural language processing, the bank enables users to perform transactions and access services seamlessly to provide a more efficient and personalized banking experience.

Meanwhile, in the United States, the Bank of America’s virtual assistant “Erica” has surpassed 2 billion interactions since its launch in 2018. It has helped 42 million clients with “personalized insights and guidance,” including assisting clients in understanding spending behaviors, providing account numbers or routing numbers when asked, finding transactions, and even aiding with money transfers and bill pay.

A New Set of Risks Rises with Conversational Chatbots

Several ethical concerns surrounding the usage of AI have been widely identified, including racial and gender biases and data privacy violations. However, conversational AI can introduce even more risks as these tools aim to imitate how humans talk and connect emotionally.

Modern chatbots are powered by (GenAI), an artificial intelligence technology that can produce various types of synthetic data. These GenAI tools present significant risks, including the potential for misinforming the user by creating realistic yet inaccurate content. For example, after a chatbot provided incorrect information about an airplane ticket discount, a court ruled that the airline was liable for the misleading information.

We must make sure AI delivers on its promises to women customers. Research shows creditworthy women may be more likely to be denied credit than creditworthy men when using automated credit scoring systems. AI-driven recommendations or ads may inadvertently reinforce gender norms and limit opportunities for women (e.g., offering less lucrative job ads or lower credit limits). These biases are not always visible, but they are significant challenges to pursuing fairness through machine learning and artificial intelligence.

A Set of Principles to Develop Trustworthy Digital Payments

As digital financial services rapidly integrate GenAI into their products, a key step in keeping trust is to inform users about the implementation and use of AI clearly. Transparency, accountability, and explainability (how an AI system works) are essential, as there is a risk of malicious use by industry outliers. For example, it has been suggested that a conversational AI could be used to upsell financial products to consumers; trust could be hurt if this is done without fully informing them of the conditions of the product or even by being emotionally manipulative.

That is why the UN Principles for Responsible Digital Payments are more relevant than ever, particularly Principle 6, Be Transparent. Financial products often come with complex contracts and terms of use. Simple wording should be used to describe features, including the scope and limits of their AI Chatbots, privacy policies, fees, transaction limits, exchange rates, and actions if unauthorized transactions occur. Messages must be brief, in local languages, and visually appealing or voice-activated to ensure comprehension, especially for low-literacy users.

With conversational chatbots, users of financial services must clearly understand whether they are interacting with Artificial Intelligence or a human, how the information they share in their interactions can be used to enhance the AI models that power it, and what information is shared with and used by third parties. There should also be precise grievance mechanisms for AI-related issues.

The Better Than Cash Alliance is actively collaborating with members and partners to highlight innovative and responsible uses of AI in digital financial services, highlighting emerging good practices, industry rules, and regulatory policies shaping the financial sector. Our aim is to contribute meaningfully to the global dialogue on the governance of AI, in line with the recommendations outlined in the United Nations Secretary-General’s High-level Advisory Body on AI’s final report, Governing AI for Humanity.