Conversational AI In Banking: Real-World Use Cases and Challenges

Conversational AI in banking holds the promise of transforming customer experiences by reducing Customer Effort Score (CES) while boosting Customer Lifetime Value (CLV) and Net Promoter Score (NPS). A clear win… on paper. Yet, for many banks, the road from potential to payoff is riddled with obstacles.

Capgemini highlights that 39% of banking institutions can’t get AI right and end up dissatisfied with adoption outcomes. In our guide, we’ll show you how to move beyond experimenting with conversational AI for banks to implementing it at scale — unlocking its capabilities and delivering real, lasting value. 

The evolution of human-free communication within the banking sector: from reactive servicing to proactive conversations 

The history of conversational tools used by financial institutions starts with inglorious automated voicemails, that left customers more annoyed than satisfied. But how far have we come since those days? Let’s dive through the three major eras of conversational banking:

• Traditional script-based chatbots and automated voice assistants marked the rise of conversational banking. While they were supposed to make dealing with customer requests easier, their rule-based nature often left users frustrated and seeking help from customer service representatives. Those early solutions were characterized by inefficient communication, a lack of memory for past interactions, and limited personalization value.   
  
• AI chatbots backed by machine learning algorithms made a breakthrough in addressing basic queries without involving human agents. Along with saving time (and headaches) for both customers and bank staff, AI-driven chatbots excel at gathering customer data. By analyzing user needs, spending habits, and behavior patterns, banks can level up personalization of their offerings and services. However, these chatbots serve mainly as trusted information sources and can’t act on the customer’s behalf.
  
• Intelligent virtual assistants became the next frontier of banking conversational AI. Akin to large action models (LAM) that have been gaining momentum since 2024, their capabilities go beyond understanding natural language queries and providing instant, relevant responses. Powered by multiagent architectures, virtual assistants can execute tasks for users, like transferring funds, creating savings accounts, and setting up investments. With machine learning at their core, they deliver hyper-personalized experiences, proactively offering suggestions and solutions that align with individual financial goals. 

From human-like chatting to ROI: top 5 conversational AI use cases in banking 

Initially, conversational AI usage in the banking sector was limited to front-office operations, covering customer support and personalized offerings for customers. However, since generative AI hit the mainstream in late 2022, the technology gradually made it to core banking services and back-office activities. 

1. Customer onboarding 

You snooze, you lose — that’s how it works with customers who are getting harder to impress than ever. For banking institutions’ online services, customer engagement and retention are pressing challenges, especially since banks naturally lag behind other industries like, let’s say, ecommerce, where the average visit to an app lasts twice as long as to a banking app. 

Therefore, rethinking interactions with users at every touchpoint is vital, and conversational AI can improve your statistics. Harnessing the technology to guide users through the onboarding process is one of the scenarios. Be it opening a first bank account for a B2C user or registering an e-signature for a B2B customer, an AI-driven bot or virtual assistant ensures a smooth experience by: 

• Requesting IDs for initial validity checks 
• Walking customers through document submission step-by-step or submitting the documents by itself
• Providing real-time updates on account setup progress     

2. Customer support  

First-rate customer support is another pillar for ROI-boosting user retention and building consumer loyalty. Implementing AI-powered solutions on the front line of customer interactions benefits both sides:

• Consumers get human-like, instant support 
• Banking institutions cut costs by covering more requests with automated customer service 

Statistics indicate that up to 60% of customer interactions can be seamlessly handled by digital assistants. For instance, conversational AI tools shine in areas like account management and credit card services, freeing up your app support and call center specialists from dealing with numerous trivial inquiries, such as:

• Updating account information
• Transferring funds
• Disputing transactions
• Checking credit scores
• Resetting account passwords

• Activating card
• Resetting PINs
• Reporting lost or stolen cards

The more accurate the chatbot, the higher the ROI from the technology, and this approach applies to any industry. When crafting an AI-powered customer support solution for a mobile taxi app, we achieved 97% accuracy in answers thanks to training an underlying LLM on a dataset that also included user queries with foreseeable common typos.  

3. Advisory services for personal finances

In-depth, fulfilling individualization of customer experience with the help of financial assistants ignites the growth of customer engagement in two directions:

• Longer sessions in a banking app
• More interactions with the app

In the pre-AI era, personalization was an underperforming feature. Now, it’s a product that delivers genuine, much-coveted value. 

AI-powered assistants actively decode customer behavior, predicting their needs and making relevant, proactive nudges before they even ask. For example, if a customer has a deposit for traveling, a bot can initiate a conversation, offering a timely deal on travel insurance. This way,  the customer gets insurance on favorable terms, and the bank cross-sells their partner’s products. 

Another real-world example of how financial organizations drive greater value for users comes from Bank of America. They trained their chatbot to jump in when a customer’s credit score drops, offering tailored advice to improve it. 

While AI systems provide extensive opportunities for ordinary customers, they truly shine when it comes to enhancing investment experience. For instance, chatbots and virtual assistants can analyze market events and prepare risk profiles for traders. 

4. Assistance to C-level executives

Along with personal assistants stepping up customer convenience, conversational AI tools are changing the game for C-suites. Assistants to high-level managers empower them to make informed, error-free decisions faster.

Let’s take a virtual assistant to a chief experience officer (CXO) as an example. PwC survey highlights that a third of the time in this role is spent on operations, related routine tasks, and follow-ups. In fact, up to 60% of that time goes into chasing down metrics from the management information systems team. But with an AI assistant, all those hours could be saved for focusing on strategy, not on tracking down information.   

Instead of diving into endless reports or sifting through folders on the company drive, what if the CXO could just ask the AI for the latest insights on sales, partner performance, customer profitability, market benchmarks, customer lifecycle, or even the NPS across different channels? Now that’s what we call efficiency.

5. Employee onboarding and training

Onboarding just got a whole lot easier, thanks to conversational AI. Gone are the days of employees drowning in a sea of tabs and apps to find answers. Now, new hires can get up to speed on core banking systems and processes with a simple chat — no more endless searching. The AI chatbot becomes their go-to source for all things info-related, reducing mental load and making their transition smoother.  

Besides onboarding, banking conversational AI takes employee training to a whole new level. Let’s say you have established customer personas that require different communication styles and strategies. With an educational chatbot, you can simulate interactions with all these personas to train sales and customer service staff for high-stakes conversations beyond the reach of AI-powered customer support. 

Conversational AI in action: charting a new frontier for a Czech bank 

Financial institutions aim to rewire customer services by relying on advanced data analytics and technologies such as natural language processing and AI (be it generative AI, conversational AI, or both). Our client — a next-gen Czech bank — decided to transform their traditional in-app chatbot into a powerful text- and voice-based sidekick to boost customer retention and satisfaction. 

Instinctools’ team deployed a private instance of GPT-4 and worked on two features with different levels of access to banking and customer data:

1. By default, the chatbot has access only to a sanitized dataset of internal banking data, such as answers to the FAQs, information about bank offerings, instructions for performing various operations, etc. It’s enough to guide customers through basic card management, transactions, insurance claims, etc. 

2. When the chatbot is given explicit customer consent to access some of their profile data, it turns into a full-scale personalized financial advisor ready to proactively help users and provide tailored insights on any banking topic. 

Risk it for a biscuit… But is the biscuit worth it? 

Can conversational AI deliver much-coveted ROI? A closer look into possible financial and operational benefits, backed by Deloitte and McKinsey surveys, indicates the benchmarks to look up to:  

• Up to 35% increase in front-office staff productivity
• Up to 15% improvement in the cost-income ratio over the five after conversational AI adoption
• 40% to 50% reduction in service interactions
• 20% to 30% lower incident rate 
• 20% reduction in cost-to-serve

Tech foundation and challenges of adopting conversational AI in banking: remedies provided 

As you see, the rewards of implementing conversational AI are high. But so are the risks. You cannot magic away challenges such as source code deficiency, data security issues, LLMs’ bias, limited visibility into the AI system’s function, AI privacy concerns, inadequate scalability of legacy software, intellectual property violations, or maintenance difficulties. However, recognizing the perils upfront makes dealing with them easier. 

The core of most of these hurdles boils down to three pillars of software development: data, cloud, and APIs. Rewarding conversational AI adoption is off the table while this bottom line isn’t covered. 

Without a strong foundation, adopting conversational AI systems is like trying to build a car from mismatched, rusted parts — it just won’t run.

The good news is that the future of your solution is yours to shape: 

• Data. Your AI engine is only as good as the data it’s trained on. Therefore, clean, comprehensive, and bias-free data is fundamental when it comes to crafting an accurate and trustworthy AI solution. Prioritize top-notch data management to create a single source of truth and provide role-based access that empowers every team member, from entry-level employees to the C-suite.
• Cloud. There’s a reason why companies with the highest profit margins are the ones with 30+% of their workloads running in the cloud infrastructure. The resilience, scalability, and budget savings cloud computing offers are too enticing to ignore. 
  Imagine being able to set up a new environment for your AI-driven chatbot or assistant in minutes instead of days and how it may speed up time to market for your software. Not to mention cloud automation and the ease of maintenance when it’s delegated to a trusted cloud implementation partner. 
• API. Well-documented APIs are easy to use and empower banks to seamlessly integrate conversational AI tools with their other products. 

When the baseline is covered, make sure to address other important aspects of your risk management plan. For instance, adopting a responsible AI (RAI) framework is one of the best practices for safeguarding your AI-powered banking software. This approach spans over six risk categories — put all of them on the front burner when implementing conversational AI. 

• Set up a human feedback mechanism for reviewing automated decisions to ensure fruitful human-machine collaboration.
• Keep documentation on implemented conversational AI tools in order to make their usage transparent and traceable.
• Source and scrutinize training data properly and adopt a mechanism like Reinforcement Learning with Human Feedback (RLHF) to wipe out the probability of biased outcomes.
• Safeguard end-user confidentiality by separating sensitive information from public data and anonymizing and/or encrypting it to ensure top-level privacy.
• Organize your AI computational resources the way to impact the environment as little as possible. 

Don’t miss the chance to hop on the conversational AI express – get your ticket to the future of banking 

The era of conversational AI in banking is here, and it’s moving fast. If you want to keep up and be truly customer-oriented, you cannot opt out of it. 

However, conversational AI isn’t a simple plug-and-play technology. You need subject matter experts with battle-proven experience to hit it big with a next-gen chatbot or digital assistant.  

FAQ