Embracing GenAI: A Roadmap for NZ Financial Services Companies - Part 2

With artificial intelligence (AI) becoming integral to everyday work, NZ Financial services leaders must reconsider how they manage people, processes, and projects.

Friday, August 9th 2024, 5:47AM

by Clive Fernandes

In the first part of this series, we discussed reorganising work around generative AI. Here, we explore how introducing friction in AI usage can be beneficial, particularly in reducing hallucinations and increasing accuracy for financial services use cases.

Beneficial Friction in AI Usage

Recent research by MIT Sloan’s Renée Richardson Gosline and doctoral candidate Haiwen Li indicates that nudges can help users spot errors in AI-generated text with minimal impact on efficiency. In a study with Accenture, participants wrote executive summaries using ChatGPT-generated text. Groups that received nudges in the form of the highlighted text identified more errors and omissions. These nudges create "beneficial friction," improving the accuracy of AI outputs.

Key takeaways from the study include:

User Input: Consider the input (prompt) into the AI tool, given users' tendency to accept up to 80% of the model’s output.
Error Visibility: Making errors more conspicuous can reduce overconfidence in spotting mistakes.
Human Reactions: Experimentation is crucial to understanding human reactions to machine learning outputs.

This insight is particularly relevant for NZ financial services companies. AI can be used to draft regulatory compliance documents or investment reports, but human oversight is crucial to ensure accuracy and adherence to local regulations. By implementing beneficial friction, such as highlighting potential errors, companies can enhance the reliability of AI-generated outputs while maintaining compliance with New Zealand’s stringent financial regulations.

Finding Beneficial Friction

Software projects in the past have been all about removing friction, so it is paramount to ensure we understand the difference between good and bad friction.

Renée talks about good friction in another article. What is good friction, and how can you differentiate it from bad? Good friction is a touch point along the journey to a goal that gives humans the agency and autonomy to improve choice rather than automating the humans out of decision-making. This approach is decidedly human-first. It allows for reasonable consideration of choices for the user, and gives the management team a clear understanding of the implications of choices.

Beneficial friction can play a vital role in enhancing the deployment and usage of AI technologies in New Zealand financial services. An example of how beneficial friction can be integrated is below.

A KiwiSaver provider in New Zealand decides to implement an AI-driven investment advisory tool designed to offer personalised investment recommendations to its clients. Instead of fully automating the decision-making process, the provider incorporates beneficial friction to improve user engagement and trust.

Interactive Decision Points

The AI tool presents clients with a range of investment options, highlighting each choice's potential risks and returns. Instead of making the final decision for the client, the tool encourages users to explore different scenarios by adjusting variables such as risk tolerance and investment horizon. This interaction allows clients to understand the trade-offs and make informed decisions based on their financial goals.

Educational Insights

At each decision point, the AI tool provides educational insights and contextual information to help clients better understand the implications of their choices. For example, it might explain how changes in interest rates could impact their investment portfolio. This educational component empowers users to learn and grow in their financial literacy.

Feedback and Customisation

Clients are encouraged to provide feedback on the AI tool’s recommendations, allowing the provider to refine and customise the tool based on user preferences and behaviour. This feedback loop fosters a sense of co-creation, as clients feel their input is valued and contributes to the tool’s ongoing improvement.

Introducing beneficial friction in AI usage provides a valuable strategy for NZ financial services to enhance accuracy and build trust with their clients. By allowing users to actively engage in the decision-making process and encouraging thoughtful consideration of options, financial institutions can create a more transparent and user-friendly experience. Beneficial friction empowers clients with knowledge and autonomy, ultimately fostering stronger relationships and more informed financial decisions.

Understanding the delicate balance between automation and human oversight will be crucial as AI evolves. Financial services companies that incorporate these principles into their AI strategies can ensure their tools are efficient but also reliable and client-centric.

In the final part of this series, we will explore how shifting the focus of AI implementation from governance to enablement can further enhance innovation and adaptability in the financial services sector. This approach will discuss how organisations can empower their teams to experiment and develop AI solutions that meet the dynamic needs of their clients and the industry by creating a culture that prioritises enablement and agility in the face of rapid technological advancement.

Clive Fernandes is a director at National Capital.

Tags: AI