Engaging Customers with Chatbots: A Guide in the Age of AI-powered Service

Photo by Mariia Shalabaieva on Unsplash

The customer service landscape is undergoing a significant transformation. Gone are the days of lengthy hold times and faceless interactions. Today, a new wave of virtual assistants, or chatbots, is taking centre stage.

These AI-powered chat interfaces are fundamentally changing how businesses interact with their customers. But what exactly are chatbots, and how can they be effectively used to create a seamless and engaging customer experience? This guide delves into the world of chatbots, exploring the strategic considerations for crafting a successful customer service strategy.

Chatbots, NLP, and Strategic Considerations

A chatbot is a computer program that mimics human dialogue. Depending on the technology used in building one, a chatbot can do much of anything from answering frequently asked questions to diagnosing and proffering custom solutions to your problems as a business customer.

Today, a common technique fueling these conversational computer programs is called Natural language processing (NLP). Natural Language Processing is a technique — -more accurately, a field — -within the Artificial intelligence toolkit that makes it possible for computers to learn, manipulate, and create stuff using human languages like English, Igbo, Latin, etc.

NLP makes it possible for you to interact with computers as you would with another human; text or voice. More than just knowing the definitions and usage of the words in human language, it enables otherwise impersonal machines to understand the underlying intent and even sentiment, behind your message. And to crown it all, Natural language processing equips these machines to respond in ways similar to how a regular human would, albeit with better grammar.

Modern NLP relies on the most prominent technique in Artificial intelligence today: Machine Learning. Some of the most powerful instances of NLP referred to as Large Language Models, think ChatGPT-4, Gemini, and Claude, are mostly built around a subset of machine learning called deep learning.

As the name, Large Language Models, implies, this new generation of NLP has learned how to use human language from very extensive data (virtually everything ever posted online, some say). These Large Language Models (LLM) are heralding the end of an era for declarative chatbots — -those built around rules and automated responses. They can imitate a variety of conversational styles, and as their umbrella name, generative AI, suggests, they can use whatever customer service knowledge base that you show them to produce responses and solutions to your customers in more ways than you can think of on the spot. And more importantly, building chatbots around LLMs is super convenient.

But before you go about hiring a developer (or start thumping your keyboard yourself, yes, some GenerativeAI providers offer low-code and no-code environments you can wrap your head around in half an hour), here are a few things you might want to have in place first:

• A clean, consistent, detailed knowledge base: Remove all contradictory, ambiguous, and superfluous details from your customer service knowledge base. Make sure it covers frequently asked questions and then some more.GenerativeAI generates stuff and you wouldn’t want it to be stuff that leaves your customers more bewildered than before.
• Seamless User experience: If you can’t hire a UX(user experience) specialist, or a developer with a strong UX intuition, you are better off using a GenerativeAI provider that offers something called an “application layer” Essentially, this means that they offer a nice User interface via which your clients can interact with your super chatbot.
• Boundaries: Generative AI can, well, generate. You wouldn’t want your chatbot saying “O ma se oh” (a Yoruba word meaning “It is a pity”) when your customer makes a complaint. It is a short walk from it joining the customer to berate you and your less than ideal product/service. Consult with your GenerativeAI service provider to see how some sort of seamless handover to a human customer service representative can be implemented when your super bot is feeling under the weather. With a well-prepared knowledge base, you will be surprised how infrequent such handovers would be, given the immense capabilities of these large language models.
• Clearly Defined Purpose: Identify the specific tasks and goals you want your chatbot to achieve. Is it meant to answer frequently asked questions, collect basic information, or provide preliminary troubleshooting steps? A well-defined purpose ensures the chatbot remains focused and delivers value within its designated scope.
• Training with Real Customer Data: The quality of a chatbot’s responses hinges on the data it’s trained on. As good as natural language processing has become, feeding the chatbot with a vast amount of your own real customer data, including past interactions, common questions, and natural language variations, allows it to develop a comprehensive understanding of your customer queries and communication patterns.
• Emphasising Transparency: Be upfront about the limitations of the chatbot. Let customers know when they can expect assistance from a human agent and provide clear pathways for escalation. This fosters trust and prevents frustration in situations where the chatbot cannot fully address the customer’s needs.
• Continual Learning and Improvement: Chatbots are not static entities. Regularly monitor their performance, analyse user interactions, and gather feedback to identify areas for improvement. This ongoing process allows the chatbot to learn and adapt over time, enhancing its ability to effectively serve customers.

Benefits and imitations of Chatbots

Building a successful chatbot goes beyond simply employing the latest AI technology. It’s crucial to have a deep understanding of the strengths and limitations of these digital assistants. If you have an online business you’re probably already aware of the benefits of chatbots to your brand:

• 24/7 Availability: Providing Unwavering Support: Unlike human agents who require breaks and vacations, chatbots are tireless workers. Customers can access immediate support, anytime, anywhere, regardless of the hour or day. This ensures that businesses can address customer inquiries promptly. With GenerativeAI, this support goes beyond simple standard responses to deep diagnosis and even resolutions.
• Faster Resolution: Streamlining Simple Inquiries: For common questions and straightforward issues, chatbots can provide quick and accurate answers, saving valuable time for both customers and businesses. Imagine a scenario where a customer has a question about their order status or needs to reset their password. A well-trained chatbot can efficiently handle these inquiries, freeing up human agents to focus on more complex customer interactions. And a well-trained Generative Chatbot can up the ante on what can be resolved without a human in the loop.
• Reduced Costs: Optimising Resource Allocation: Chatbots can handle a high volume of inquiries, particularly those related to frequently asked questions or basic troubleshooting. In my experience, these sorts of inquiries constitute the lion’s share of the queries to customer service. This translates to significant cost savings for businesses. By deflecting these routine inquiries, human agents are available to address more intricate customer issues that require a significant level of specialised expertise.

However, as you also probably know, chatbots are quite limited. And if you have ever used ChatGPT or Gemini extensively, you would also be aware that despite their impressive abilities to fool a human into thinking that they are conversing with another human, sometimes, they seem to just lose their minds in the most cocksure manner. There are quite a few technical reasons for this:

• Hallucinations: While LLMs are trained on a very, very, large corpus of data, and while in most cases, these would be enough, it can’t be complete. There will always be some data left out, be it current info or just rare circumstances or issues unique to a customer. But, commonly, GenerativeAI programs don’t bother to check or even think of acknowledging that this falls outside their area of true expertise. Instead, they proffer their own theories and solutions, entirely made up from thin air. And this can be more than just a minor irritation, both to your customer and to your business.
• Limited Understanding of Human Language:: Generative AI and the chatbots they now power are astonishingly good at using human languages. But there’s a small, though, important difference in using language and truly understanding the communicative intent behind the words used. A human might ask follow-up questions about what he/she doesn’t get. Chatbots powered by GenerativeAI might not unless purposely instructed to do so.
• Frustration Factor: Stemming from the two above, this frustration factor. I have used enough generative AI tools to know of this well enough. Reasoning is not one of their strong suits. At least a rule-based chatbot would readily indicate when it has reached the limits of its ability, LLM based chatbots would rather tell you their faulty responses. They really have no sense of how concepts actually play out in the real world, especially when these are used in ways unfamiliar to them.

In any case, all these limitations show up infrequently, and keeping in mind that a clean, consistent, and detailed knowledge base is the foundation of awesome chatbot performance will make instances of these rarer still.

Chatbots are a valuable tool for enhancing customer service, but it’s important to remember that they are just one piece of the puzzle. The ideal scenario involves a strategic combination of chatbot assistance and human expertise. Think of it like a well-rehearsed play. The chatbot acts as the opening act, efficiently resolving basic issues and handling the initial wave of customer inquiries. When the situation becomes more complex, or the conversation requires a human touch, the chatbot seamlessly hands the baton over to a live agent. This agent is then equipped with the context of the previous chatbot interaction, allowing for a smooth transition and personalised service.

As AI technology continues to evolve, chatbots will undoubtedly play an increasingly prominent role in shaping the future of customer service. For the foreseeable future, however, don’t fire all those sweet-voiced ladies( or Gupta) just yet.