Multiple researches show that chatbots are relatively over-hyped in the marketplace and shown in the press to make banks appear more modern and convenient to customers. In reality, most chatbots in banking are pilot projects with little or no evidence of ROI. Risk-related banking functions such as fraud detection are easier use-cases of AI than conversational interfaces that require advanced natural language processing (NLP) algorithms. Banks might want to show off their chatbots but that is not where they are investing.
Given the high-speed developments in artificial intelligence (AI) chatbot technology and its implementation, it is now possible for chatbots to both improve customer satisfaction and a company’s financials, with a proper implementation plan though. In this article we will discuss some of the challenges and chatbot strategy that will help leaders think realistically.
Chatbot implementation plan
To improve the customer satisfaction, chatbots must be able to interpret each exchange in a dialog in the context of conversation. Chatbots have to be trained to understand the specific terminology and phrasing of the user community. Along with customer service improvement, it is important to have a clear idea of the types of ROI you are looking to deliver with your chat or conversational system.
It is important for leaders to have a complete understanding of the capabilities and potentials of chatbots in banking and challenges of deploying them. Here we will break down some of the objectives in a way that would be convenient and useful and get a couple of different types of ROI for business leaders.
Improving customer response time is one measurable benchmark that businesses might use for an AI system. It is possible to have a chatbot that can answer rote questions that retail banking customers ask. For example, “What is my account balance?”, “How to reset a password?”.
In order to get the optimal returns on investment, a cross-functional AI team within the bank would have to list the most repetitive questions asked by customers through chats and voice messages and train a machine to respond to bounded and understandable questions.
The questions can be divided in 3 sections:
(a) Technically possible to respond to with AI
(b) Valuable to users
(C) Valuable to the business
For example, users can ask a question regarding balance check in multiple ways. Identifying the variations on a theme and adding them in the database of a chatbot application can help meet the challenge. Understanding the most frequent and most boundable questions will help for faster and more effective responses.
Chatbots can be trained to handle some customer transactions more quickly and accurately than human agents, but only if these transactions are understandable to chatbots. For this, it requires training chatbots to be able to get the right updated message across.
Self-reported customer satisfaction
Many companies allow their customer to rank their customer service experience, for example on a 1-5 scale of satisfaction, to know how satisfied they are with customer support services. This benchmark of satisfaction can be experimented with a chatbot in banking.
Being able to learn about what customers like and what is upsetting them through chatbot will help financial institutions perform better. Not only will it help banks actually tackle questions, but natural language processing (NLP) can be used on the backend. Banks can learn from chatbot conversation on how to improve customer experience.
Another objective of deployment of chatbot would be cost reduction. Using AI-powered chatbot to handle customer transactions would free up human agents to do other kinds of value-added and more productive tasks within the bank, such as sales calls. This would reduce the need for hiring more employees for overseas call center, ultimately reducing the overall outsourced services expenses. Employment of a chatbot might effectively grow the customer service function with headcount and payroll costs.
At the onset of a chatbot implementation, a bank would need to figure out what cost savings metric or benchmark it wants to adhere to. For example:
- Do we know how many outsourced agents are needed when our customer base grows by X%, and can a conversational interface effectively replace that headcount increase?
- Is it possible for the bank to quantify the reduced FTEs on customer interactions, and to attribute that to staff reduction in call center or the customer function service as a whole?
Implementation of chatbots to increase revenue is also a viable goal in financial services. Banks can use the conversational data to identify customers at churn risk and take actions to retain them by giving them special offers through email or text messages. Chatbot applications are helpful in cross-selling and up-selling. Chatbot can identify customers with the highest likelihood to purchase additional products or services and pass those customers to sales persons who can handle their needs and sell them what the user is looking to buy.
Banks can compare the success rate of closing the deal in opening a business account or loan when routed through a chatbot first, or to a human agent first. It is a good way to estimate the ROI.
Building a cross-functional AI execution team
To assess viable use-cases for chatbot and viable measures of success, some degree of cross-functional team will need to be involved. The specific combination of cross-functional team will vary from project to project.
- Dedicated subject matter experts: It involves customer service specialists who understand the customer’s needs, conversational flows of the company and how customer service issues impact a business.
- Dedicated data science talent: Usually data scientists don’t understand subject-matter data. To develop intelligent hypotheses, data scientists need to work shoulder to shoulder with subject matter experts who understand the technical issues related to customer service and process of the business.
- IT champion: While many full-time IT personals may need to be a part of an AI transformation project, there must be at least a champion within the IT function because AI projects typically hurl a variety of new issues and demands.
- Business champion: A busines champion who believes in the project and wants to make it successful has conceptual understanding of AI implementation and use-cases and can connect with the said decision-maker.
The most common reason for unsuccessful chatbot implementation is failing to involve dedicated subject-matter experts and IT people.
Performa data audit
It is one of the most important aspects of a chatbot implementation plan. It is advisable to carry out data audits across all of existing chat logs or call center logs to get a better understanding of the type of information available. Data audit would also involve determining the accessibility of the data in a format that is organized and understandable.
Chatbot implementation plan is a multiphase process and needs a sound strategy. It is easy to get it wrong over 90% of time for a variety of reasons. For the right implementation of conversational interface, leadership must develop a reasonable understanding of what the technology is actually capable of. Cross-functional teams must assess viable use-cases for a chatbot application in banking and realistically determine how project success should be measured.
AI technology is evolving rapidly, and many chatbot software providers are incorporating the latest advances into their products. These leading-edge chatbots provide multiple semantic processors to expand the range of their capability to understand language and add sentiment engines to be able to detect the emotional state of the user. These “emotionally aware” chatbots can determine and rank a user’s attitude along a variety of measures such as anger, disgust, fear, sadness, joy, and positivity, on both a per-exchange basis or across the entire dialog.