LugahBot: Helping Banks Build Differentiation
LugahBot is a white-label Conversational AI Chatbot solution for banks to deliver hyper-personalized digtal banking experiences via conversational chatbots. I led the design and integration of the AI conversational bot for 5 African banks, resulting in a 40% reduction in the average customer query resolution time and significantly increasing customer satisfaction.
Background
COVID-19 disrupted the banking ecosystem. Almost overnight, customers defected from their long-standing preferred traditional banks to new digital-only entrants that offered personalized and convenient banking from the comfort of home.
From retail to education, most industries went online during the pandemic. Customers expected the same agility and convenience from banks and financial institutions – an expectation that accelerated AI adoption for fintech companies and the birth of Lugahbot
LugahBot is an enterprise Conversational AI chatbot solution for finance that helps banks automate customer service, and deliver a simple, more intuitive method to bank digitally thus improving both the customer and agent experiences.
Opportunity | Building an AI-Powered Contact Center Solution and Mobile Banking Chatbot Experience
We all know that the key to banking success remains stellar customer experience. For customers they expect to be known, understood, and able to get just what they need from your institution on demand and on their terms. Moreso, customers will always have a lot of questions when their hard earned money is involved, with a lot more requiring personal assistance.
However, banks today are struggling to meet these expectations as they face increasing customer service demands from simple to complex queries. This places significant strain on customer call agents and leads to slower query resolution, and increased operational costs as banks attempt to maintain 24/7 support.
Our team at AIDML conducted preliminary industry research and the following challenges were discovered:
Several key problems were identified:
High influx of customer inquiries: This influx leads to increased customer dissatisfaction due to delayed response times and inefficient resolution processes.
Overwhelmed call centers: Banks' call centers receive a large daily influx of calls, resulting in long wait times for customers who are hoping to speak with a representative. This slows down the resolution rate for customer queries, leading to frustrated customers and significantly impacting overall customer satisfaction.
Complex bank app interfaces: The overly complex interfaces of banking apps cause app fatigue for customers. Many customers find themselves puzzled and frustrated when trying to navigate through a maze of features to find even the most basic functions.
Feature overload: Today's online banking interfaces are often overly complex due to the addition of numerous features. While intended to enhance the user experience, this complexity can make it difficult for customers to find the information they need quickly, resulting in further frustration and dissatisfaction.
Problem statement
While banks strive to deliver efficient customer service and user-friendly digital interactions, they face challenges in offering a seamless 24/7 support system and an intuitive online banking experience, leading to customer dissatisfaction and high churn rate for existing customers.
Goal
The goal was to create an AI bot solution that resolves customer queries faster, performs transactional services for customers, and delivers a personalized banking experience, ultimately boosting customer satisfaction.
Banks need a way to provide personalized, efficient, and accessible customer service round the clock tailored to each customer’s need. This way, customers get 24/7 support without in-person visits or lengthy phone calls, invariably boosting customer satisfaction (CSAT).
How might we
...help banks build differentiation, and provide personalized & timely service to customers; the "modern bank customer".
Solution | Improving Customer Satisfaction and Reducing Avg. Customer Resolution time by 40%
Build an AI Conversational bot solution to manage this influx thus, reducing wait times and operational costs by automating routine inquiries and transactions.
Why AI bots?
We realized that AI bots are ideally suited to the banking sector where customers frequently have similar, repetitive queries that can be automated, providing instant responses and reducing the load on human agents.
Industry reports by McKinsey, highlighted a:
30% customer preference for quick digital solutions over calling customer service.
In October 2021, working in a cross-functional team, I led the design direction of the project in collaboration with the ML Engineer, launched a B2B Enterprise Conversational AI Chatbot. This solution was first demonstrated at GITEX Dubai 2021. It utilizes conversational and voice-based AI to facilitate banking transactions, significantly reducing the volume of calls routed to human agents and accelerating the resolution of routine customer inquiries. For instance, customers can now check account balances, inquire about new products/services, or log transaction disputes directly through conversations with the bot, bypassing the need to wait for live agent assistance.
Following the launch, we enhanced the chatbot's functionality by integrating a seamless handoff process to human agents when the bot cannot fully resolve a query. This feature ensures that agents receive the chat log for context, allowing them to engage more effectively and swiftly resolve customer issues.
Outcome
Impact highlights:
We launched the mobile iOS and Android app versions at the GITEX Dubai 2021 Demo.
6 months post-launch, we successfully digitized customer service experience for 5 large banking and insurance companies with average customer base of 100,000+
Recorded reduced volume of routine inquiries to human agents and operational costs reduction by 30%, reducing customer response time by 2.5 minutes.
17% increase in customer satisfaction and 44% reduction in customer effort score through automation of document collection for account opening and customer onboarding
Design process
Discovering the problem-space: understanding the needs of the end-target users
The first step in the design process was to identify the need for LugahBot and understand the specific needs of bank customers. The objective was to determine the key problems we aimed to solve through AI.
Although the business model is B2B, it was paramount to comprehend not only the needs of our banking companies (our direct users) but also ensure that the product we were creating was user-centered. This meant designing a solution that bank customers would find useful and would want to use for their banking needs.
Based on our hypothesis of building a chatbot solution to address the high-level problem of customer service inefficiency, my design process began with an in-depth analysis of the "bank-customer relationship" in Nigeria.
Findings/Research Insights:
Today, customers expect a personalized experience from their financial institution…70% of customers feel frustrated with the long wait times associated with traditional customer service, leading to dissatisfied customers
Chatbots can cut down these wait times by handling multiple queries simultaneously.
Customer Expectations: Today, customers expect a personalized experience from their financial institutions. They want quick, accurate, and personalized responses to their queries.
Customer Frustrations: 70% of customers feel frustrated with the long wait times associated with traditional customer service, leading to overall dissatisfaction.
Potential of Chatbots: Chatbots have the capability to cut down these wait times by handling multiple queries simultaneously, providing timely and efficient customer support.
These insights highlighted the critical need for a solution like LugahBot, which could enhance customer satisfaction by reducing wait times and providing personalized banking experiences.
Building on research insights
To make our value proposition clear and also understand what bank customer base could be targeted for this offering, we came up with a persona based on Market segmentation: Identified key customer segments that would benefit most from LugahBot, such as tech-savvy millennials, busy professionals, and users frequently engaging with online banking, ensuring focused and relevant efforts.
The proto-persona
Modern bank customers
Tech savvy, 25 - 45 year olds that have witnessed the shift from traditional banking to digital-first financial services. They value speed, personalization, and the ability to manage finances on the go. They usually expect their banks to provide not just financial services but also help manage their financial health, in a way that’s accessible through their smartphones at any time of day.
Goals:
manage banking needs quickly and conveniently
receive reliable financial advice
personalized banking experience: relevant financial product recommendation
Further validation: in-depth interviews & concept testing
I employed a mixed-method research approach to understand the experience of customers in using bank applications and with customer service. Furthermore, I conducted concept-testing on the perception of chatbots among 5 "modern bank customer" personas. I wanted to understand what they thought of chatbots and in what way their experience with banking apps could be enhanced via chatbots.
Some questions that drove my conversations with participants include:
Experience: chatbot exceeded or failed to meet expectations?
How do you feel about using an AI bot for your banking needs?
What would make you trust a chatbot with your information?
What are major tasks users perform on their bank apps?
Banking tasks you would prefer not to use an AI bot for?
Findings
Customers desired a more intuitive and human-like interaction with digital banking services.
97%
opened their apps for four main task-oriented actions
85%
showed skepticism in sharing personal information
80%
expressed expectation mismatch in financial chatbot interactions
70%
lacked awareness in bank product offerings and services
Competitive analysis: chatbot audit
To correlate some of the stories from customer experience with existing chatbots, I did an audit of the top 3 bank chatbots in Nigeria:
Leo by UBA Bank
Ivy by Fidelity Bank
Ada by Diamond Bank
I took detailed notes of what they lacked, and what worked well for them. I also wanted to dive in further to see what users of each service were saying, and read some reviews online. All of these findings were gathered and used to generate ideas for potential features, identify gap in the existing product market and opportunity for differentiation with Lugah.
The gap
The competition’s chatbot was problem-specific and lacked in being personable and action-focused
Project definitions
Constraints
Due to the limited amount of time that we had to meet up with the product launch timeline and conference demo, we had to make some decisions regarding our priorities and the project methodology to ensure we deliver a working MVP.
One main constraint was the lack of access to banks' internal customer support system at scale.
Not enough time to conduct product market fit for internationalization. We were going to lauch in Dubai as well, but we didn't have enough insights to the Arabian fintech market
Technical constraints: limitations in Natural Language Processing capabilities: due to the timeframe and limited resources, we couldn't train the chatbot using advanced natural language processing in ensuring accurate responses in case on ambiguity in languages.
Collaborating with the internal team
Throughout the process, I worked closely with the product manager and engineering team to ensure we had a shared understanding of the project goals and problems we were solving for.
I organized various collaboration workshops with the team, which helped in identifying early potential blocks and trade-offs needed for the potential list of features I came up with.
Prioritization workshop
Following my presentation of the customer challenges and needs identified during research, I worked alongside the product manager to organize a feature prioritization workshop within the constraints of the project.
The insights from the user interviews formed the basis of potential features. Each potential feature was evaluated based on Impact and Value vs Technical Feasibility and Complexity.
User Impact: Assessed how much each feature would positively impact the user experience, considering factors like convenience, speed of service, and problem-solving ability.
Business Value: Evaluated the potential return on investment (ROI) for each feature, including its ability to reduce operational costs, attract new customers, or retain existing ones.
Guided by a two-fold approach: customer pain points and business goals, the following features were prioritised afterwards:
Transactional services: Account and payment
Onboarding: FAQs and Onboarding: account opening, product onboarding
Financial advisor: assist customers in better managing their funds.
Personalized banking: personalized advice and product recommendations
Success metrics
40% reduction in
Customer Effort Score
Successfully onboard 4 in 10 customers.
60% Conversions
60% of users that engage start and complete key tasks
25% Activation: Product adoption
Atleast 25% of engaged users yield bankable revenue stream
Design principles & Guidelines
Designing for conversation UI required a blend of visual design principles and linguistic UI. Here are some decisions that guided my design approach for: conversation, interaction and inclusive design.
Solutioning & Iterating
Planning the Chatbot information architecture and conversation pathway
Working in collaboration with the product and engineering team, we designed the chatbot's information architecture, ensuring it could effectively manage a wide range of customer inquiries. This thorough planning was essential for seamlessly integrating the chatbot into the existing bank application.
Comprehensive Mapping of Customer Queries
To begin, we conducted a detailed analysis of common customer queries and interactions. This involved collaborating with customer service teams to understand the variety and frequency of questions posed by customers. By mapping out these queries, we could identify patterns and categorize them into specific topics and subtopics. This hierarchical organization allowed us to create a robust structure that the chatbot could navigate efficiently.
User-Centered Design Approach
Understanding the end-users' needs was paramount. We incorporated user-centered design principles to ensure the chatbot's responses were not only accurate but also empathetic and engaging. This meant designing conversation pathways that were intuitive and easy to follow, providing users with clear and concise information. We also included fallback options and escalation pathways to human agents for queries that required personalized attention.
Integration with Existing Systems
Seamless integration with existing customer service frameworks was a critical aspect of our planning. The chatbot needed to access various databases and information systems to retrieve accurate and up-to-date information. We designed APIs and integration points that allowed the chatbot to pull data from these sources in real-time, ensuring that customers received reliable responses.
Scalability and Flexibility
Given the dynamic nature of customer service needs, the information architecture was designed with scalability and flexibility in mind. We created a modular structure that could be easily expanded to include new topics and updated as customer needs evolved. This approach ensured that the chatbot remained relevant and effective over time.
Conversation Pathways
We developed detailed conversation pathways for each category of inquiries. These pathways included scripted responses, decision trees, and prompts to guide the user through the conversation. Each pathway was tested and refined to ensure it provided a smooth and logical progression, minimizing the chances of user frustration or confusion.
Continuous Improvement
Finally, we established a feedback loop to continuously improve the chatbot's performance. This involved integrating post-chat feedback. Based on this, we made iterative enhancements to the information architecture and conversation pathways, ensuring the chatbot consistently met the needs of both customers and the business.
Pen and paper sketches, user flows
Once we finalized on the core features we wanted to include in the MVP, I came up with two possible user flows depending on “why” a user might use the chatbot application. This was to help us map out the optimal path for users to get them started off with whatever their need might be.
These flows were later turned to pen & paper sketches and eventually, hi-fidelity task flows. The task flows helped me to visualize the flow of dialogue for different user task scenarios, which also helped in identifying potential dead ends, communicate ideas and direction with the team.
User flows
Sketches
Task flows
Testing & iterations
Uncovering challenges, and making tradeoffs
In testing, I was looking to test for the following metrics: task completion rate, time to task completion, error rate, perceived effectiveness. Also, in making trade-offs, the team aimed to align with the project's core objectives: delivering a user-friendly, efficient, and reliable conversational AI within the constraints of the project.
We conducted an initial round of unmoderated testing with a small demography in Nigeria once the app features were completed. I recruited participants on user testing and provided a set of banking tasks I wanted the participants to carry out. I made sure to encourage participants to think out loud, to make better data-informed decisions in the next design iterations. After reviewing the test results, the following discoveries were made, which led to some iterations in the app design.
Trust: Users valued personalized experiences but were also highly sensitive about their financial data’s privacy and security. So building user trust while ensuring privacy and security of user data was paramount.
Tradeoff:
Standalone solution vs. integrated bank solution
Challenge: Balancing personalization with trust & familiarity
Takeaway: Need to increase user confidence in using the bot to drive trust and reliability.
Once we decided to go for an integrated bank solution, to instil a sense of trust in customers who already trust that their data is secure with their long standing traditional bank, the design of the UI of the chatbot conversation interface, needed to reflect this trust as well.
I had 3 UI explorations for the choice of the conversation UI:
Voice Assistant: built-in voice assistant directly from the bank app home page.
Instant messaging: a simple chat-like focused interface primarily conversational.
Bank interface-driven: a mix of conversational + bank-like feel.
Error handling and guidance: While NLP technologies have advanced significantly, they still have limitations in understanding context, ambiguity, and the nuances of human language. Designing LugahBot involved working within these limitations to ensure reliable and accurate user interactions.
Tradeoff:
We decided to initially forgo some advanced financial advisory features in favor of ensuring a robust foundational service for FAQs and account management, planning to introduce more complex services in subsequent updates."
Challenge: Difficult navigation and error handling
Key Insight: high abandonment during complex transaction flows and frequent repeated queries
Takeaway: Need to help users get to the desired information and to escalate issues to a human agent when necessary. Turn frustration into engaging conversations.
Iterations made
Quick reply buttons and clear prompts: designed quick-reply buttons, shortcuts for common tasks and clearer navigation prompts.
offer suggestions, ask clarifying questions, or guide users back to the main menu when it couldn’t understand a query.Fallback Options: implemented fallback options, including the ability to restart the conversation, access a help menu, or connect with a human agent, ensuring users always had a way to accomplish their tasks.
We also included a disclaimer to guide user expectations of what the bot can and cannot do.
Balancing automation with human touch: Automated responses ensure quick and scalable service but may not suffice for complex or sensitive issues. Introducing a human handoff option improved service quality for challenging queries but required additional resources and could interrupt the seamless automated experience so we opted for a phased roll-out.
Trade-off:
Automated vs Human-assisted bots
We released the first version solely based on automated bots with plans of implementing the handshake between the bot and customer agents in the next roll-out.
Final designs
We delivered a Mobile app and an Enterprise web application that banks can use to train the bots, based on RASA technology.
Empathy for users played a crucial role in shaping the designs for LugahBot, particularly in creating a more inclusive and supportive user experience. The goal was to make the conversations feel humane and engaging, in an intuitive manner.
To simplify the interaction process, LugahBot utilized quick replies and predictive text suggestions, allowing users to choose from pre-defined options instead of typing out responses, which was particularly helpful for users who were slow typists or less confident in formulating responses.
Mobile screens
Enterprise web application
Reflection & Learnings
Leading a small product design team and product strategy for a completely new area of design, though started out challenging, was an enriching experience for me. It helped me approach design from a holistic view, ensuring that I cover all bases. This project not only refined my leadership and design skills but made me a better communicator and collaborator with cross-functional teams.
Other projects
MyHealth: patient information portal
Reimagine the patient health learning experience for Canadian patients going through medical procedures through access to personalized information, empowering them as active participants.
XCel Payment App Redesign
Redesigned the XCel mobile app to drive user retention and enhance the user experience, enabling individuals and businesses make international financial transactions across 5 countries.