Many banks have already implemented AI chatbots in their customer service operations, but they were not always well received by customers. Henning Holter, Business Development Director at Star, addresses whether further implementation of AI technologies, including generative AI, can help banks to improve their customer service or if their effectiveness will wane as customer problems become more complex.
The 21st century is witnessing a remarkable transformation in technology and innovation with the advent of Generative AI. This revolutionary technology can not only democratise access to unmatched productivity for data and process intensive financial organisations, but also potentially iron out the kinks in early versions of AI such as chatbots.
Banks introduced chatbots to enhance the effectiveness of dealing with multiple customer enquiries, but the reality was often the opposite. Some people experience significant negative outcomes due to the technical limitations of their functionality. In fact more than half of UK banking customers surveyed by Capterra in 2022 claimed chatbots “never understand their requests or needs” – not “occasionally misunderstood” or “often misunderstood”; “never”. On top of this, 50% noted chatbots generated frustration by being repetitive and running in circles – potentially making customer retention less likely.
In the US, close to seven in 10 (68%) U.S. consumers have utilised chatbots for customer service, but nearly 8 in 10 (77%) prefer interacting with a human, according to a recent Ipsos poll. Further, they are more likely than British customers to report them as frustrating (77%) and 88% would rather speak to a person. The Consumer Financial Protection Bureau (CFPB) even issued a warning after receiving numerous complaints from customers claiming that chatbots had failed to provide timely, straightforward answers to their questions.
Financial institutions that adopt client-facing generative AI programs will be keen to avoid these issues as they want to enhance rather than destroy the user experience.
Benefits not limited to operational efficiencies
By automating traditional financial processes, that tend to be slow and highly manual, banks and other financial institutions can reduce the potential for friction and overall improve the customer experience.
Of course, the benefits of Generative AI are not limited to operational efficiencies gained by automating numerous repetitive or routine workloads.
Deployed effectively, generative AI is able to adjust to different levels of financial and digital literacy and modify its output to better personalise responses for both support agents and the users they serve.
It can remove standardised tasks like account creation and management touchpoints for support agents, who are then free to focus on the individual client needs, and address specialist problems like debt recovery and insurance claims. It can also empower employees by providing information to make decisions further down the hierarchy, such as making credit decisions directly from the front-line.
Further, banks can also help their customers understand relevant trends in their transaction history, proactively suggest suitable changes to credit card-, savings- or borrowing plans, and generally help customers become more financially literate in the process.
Where support is concerned, application processes can deliver timely product recommendations based on account data, credit history, and live interactions. While this has been frequent in already established AI, there’s an expectation that generative AI will only enhance the personalisation process.
Of course fraud detection and the ability to enhance customer verification processes by creating unique security prompts for customers to use while logging in can be as much a benefit for customers as for the financial organisation.
The pitfalls of generative AI
It’s important to note that generative AI has its limitations, risk factors, and a best practice methodology for adoption.
Firstly, generative AI models will not fully replace financial advisors since AI can’t fully capture the nuances of financial services or market conditions. For instance, a client may still see value in a company despite trading algorithms predicting a fall in its stock price. While trust can be built through the combined use of generative AI support and human interaction, it can be undone by system failures or machine inaccuracies.
Secondly, training data sets can create performance issues through biased results. Therefore, ongoing supervision, guidance, and feedback are vital to accelerate digital journeys and enhance employees’ performance and customer trust. Similarly, generative AI models are prone to hallucinations (where the program ‘makes up’ facts or sources). As a solution, limiting your model’s ‘temperature’ (how ‘creative’ it is) can help mitigate this and ensure that what is received is wholly accurate information.
Finally, business leaders must always protect proprietary company information and customer data. External collaboration with experienced third-partiescan ensure that it is adopted securely while also accessing expert implementation support at the same time.
Improving customer trust and satisfaction
Customer engagement and the user experience have become crucial ways for financial organisations to differentiate themselves. They can improve profitability by avoiding prospective customers dropping off in the final stages of the application process because of poor UX, which is costly.
By following these five key guidelines you can avoid some of the most common mistakes that impact the user experience:
- Close cooperation between the risk department and application developers. Resolving perceived opposing interests is key to making this work – often cultural obstacles need to be addressed to reach a shared understanding of the endgame.
- Empower staff to make decisions. Use AI to filter out the routine questions, and provide front-line staff with the tools they need to help clients faster and directly.
- Sufficient data collection points. Design the application with sufficient data collection points so that you avoid bias which can lead to discriminatory outcomes.
- Take into consideration transaction speed. Failure to do this can compromise risk thresholds. Speed should not lead to an increase in risks.
- Consider the trade-offs. Some trade offs will need to be made such as between a streamlined and friction free onboarding process and capturing sufficient information when it comes to credit or loan applications.
Resolving issues quickly and efficiently is the end goal. The removal of both digital and human touch points throughout application processes can make for frictionless service, but it is important to provide fast access to human intervention when necessary and where it adds personalised value to individual customers.
The result: customer trust and satisfaction.
About the author
Henning Holter is a financial services executive specialising in digital transformation and innovation within the banking sector. His background includes 15 years in senior leadership positions, spearheading numerous successful digital transformation initiatives that empowered organisations to harness cutting-edge technologies to deliver enhanced customer experiences. Henning’s expertise lies in leveraging innovative solutions to drive operational efficiency, optimise business processes, and foster sustainable growth.
