From mediocre chatbot to good help: the step towards conversational AI 2.0

This blog is also published on Frankwatching.com

More and more companies are using chatbots to handle customer questions quickly and efficiently. The goal is clear: to help your customer directly and reduce the pressure on customer service. The current results are mixed. Many chatbots do not deliver the experience that your customer expects. Public opinion is therefore often critical.

Yet research and practice show that customers value speed and convenience. They not only want information, but also immediate help. This may be done digitally, as long as the service level is equal to what your employees provide. And that this does not block the path to your employees. It must be a choice.

At the level where the chatbot really offers customers this kind of valuable service, only a few chatbots are active. That is why customers still value contact with an employee so much in many cases. But with the arrival of Large Language Models and generative AI, you can make huge steps in customer experience and make the leap to conversational AI 2.0. But how do you do this?

Build on the best experiences

In the Netherlands, there are already great examples at banks, energy companies, retailers and insurers. They show that with a good digital dialogue in direct contact with the customer (via voice and chatbots), you save many valuable hours and the customers experience top service.

What do these frontrunners do differently?

• They continuously invest time and energy in their chatbot
• They enable customers to not only ask questions, but also to directly arrange matters via the chatbot
• They are constantly looking for ways to improve their chatbot and align it with their customer's preferred channels
• They use a chatbot as part of their overall vision on customer contact and let it work together with the people and systems in the organization
  These companies are now building further with Generative AI to further increase customer experience and business value.

The real value is in the process

With a Large Language Model (LLM), a form of generative AI trained on your company information, your customer can quickly and pleasantly receive an answer to his or her question. But you need more. Behind every customer question often lies an action: reporting damage, blocking a bank card or changing an installment amount. Your chatbot should therefore not only be able to speak well, but above all help your customer to arrange things.

By connecting the chatbot to customer data and business processes, a new dimension in customer interaction is created. Customers can then independently:

• Take out or change insurance immediately
• Request and receive status updates
• Edit data
• Troubleshoot issues such as password resets or processing returns

These are examples of tasks that companies spend a lot of time on every year. A critical question: can't customers just arrange these kinds of things in the 'my' environment or in our app? The reality is that many customers hardly use these existing digital channels.

Successful organizations are therefore already using their chatbot as an additional channel, so that more customers opt for digital self-service. This ensures fewer phone calls, lower costs and higher customer satisfaction. The use of an LLM can make this even better.

Why Generative AI Makes the Difference

It’s all about customer experience. A bad conversation with your digital colleague immediately reflects negatively on your company. With the advent of generative AI, the playing field has changed. Consumers and employees increasingly chat with LLMs such as ChatGPT or Perplexity. These models have natural, fluid conversations and understand customer intentions much better than traditional chatbots. If your chatbot can’t match that, there’s a good chance your customer will be disappointed.

Companies are therefore now investigating how they can use the power of LLMs for their digital colleagues. Because the better your chatbot understands the customer, the faster and more adequately your customer will be helped.

From chatbot to process assistant: conversational AI 2.0

The first generation of chatbots mainly answered frequently asked questions. But your customer wants to take care of things: request a return, postpone a payment or change an appointment. Conversational AI 2.0 means that your chatbot turns into an intelligent process assistant, by combining three elements:

• LLMs for Natural Interaction and Intent Recognition
• Orchestration of customer contact and deep integration with backend systems (such as CRM/ERP for data and workflow links) to be able to handle matters
• An up-to-date, validated knowledge base based on the chatbot always using the correct information in the dialogue with the customer

The optimal combination: LLM + process bots

The real power lies in the combination of an LLM with process bots. The LLM provides a natural conversation, the process bot handles the entire process, for control and predictability in the execution.

This allows your chatbot to:

• Perform complex processes quickly and without errors
• Combining natural interactions with practical solutions
• Making customer interaction smoother and more efficient

This way you not only improve the customer experience, but you also maximize the value of every interaction.

AI Agents vs. the Hybrid Approach

The latest development in the AI ​​market is about AI agents, which is being fully deployed by tech platforms. An AI agent is a fully autonomous AI assistant that independently initiates conversations with customers and can decide for itself what the best follow-up actions should be. Fully autonomous AI agents are powerful, but still entail risks in terms of control, compliance and predictability. This all has yet to crystallize. Especially in direct customer contact, these risks are often unacceptable.

The hybrid approach, LLMs combined with process bots, offers a proven solution for this. This way you keep control, while you and your customers benefit from the use of advanced AI.

At the same time, this provides the time and space to experiment and to see whether AI agents will ultimately provide a better alternative.

The Chatbot Owner's Challenge

Either way, upgrading to a smart, integrated chatbot comes with challenges. As the owner of the chatbot, you experience increasing responsibility. Your digital colleague receives questions about your products, services and internal processes, before the customer asks his real (process-oriented) question. Errors in this immediately reflect on your digital colleague and your company.

In addition, the connection with customer data and backend processes places high security and reliability requirements. The protection of personal data and the integrity of company data must always be guaranteed.

How do you keep a grip on this? A few tips:

• Start a conversation with your colleagues. A successful digital colleague transcends the silos in your company. Work together.
• Data and knowledge need to be in order. Create a common approach that ensures that the right information is always central, up-to-date and accessible, for both your employees and your digital assistants.
• Ensure better quality of answers from your LLM. For example, integrate smart retrieval methods (read more about this in this article that DDMA posted earlier) and checks.
• Limit autonomous actions and opt for a hybrid approach.
• Implement strict security measures and ensure a robust infrastructure.
• Continuously monitor and optimize dialogues, data and knowledge based on real-time feedback.

Your next step

The time to act is now. Companies that integrate their chatbot into business processes are already achieving great results. More satisfied customers, higher efficiency, lower costs. In short: a rock-solid business case. The further development to conversational AI 2.0 will improve this even further.

* The translations on this website are provided by G-Translate from Dutch