Let's face it: don't you have a Chatbot proof of concept in your drawer somewhere in the office? It lies there, gathering dust and patiently waiting for the moment when it finally comes into action. But that moment is unlikely to come... Why do so many companies struggle to get chatbots into live production or keep them running successfully in the long run? There are actually many companies working on various Conversational AI initiatives. However, only a few are actually implemented and go into live production. And then they are often taken offline again after a short time.

There are three main problems that are probably responsible why chatbots and virtual assistants remain in the POC phase:

  • The chatbot operation or maintenance of virtual assistants
  • The interaction between departments and IT
  • The wrong choice of technology

The first problem: The chatbot operation

After the chatbot has gone live, the operation of the chatbot starts. Most companies spend a lot of time developing their chatbots or language assistants, but they rarely consider the resources and processes needed to actually run them successfully afterwards. There are four aspects that should be carefully considered before starting Conversational AI projects.

  • Establish a successful Conversational AI Team
  • Train your chatbot and make it better quickly
  • Improve your virtual assistants in the long term with a dedicated roadmap
  • Set KPIs and metrics to measure the performance of your virtual assistant

The second problem: The cooperation between departments and IT

Another critical factor why chatbots and language assistants do not go into production is the sometimes difficult relationship between the business and the company's IT department.

What usually happens is that IT selects technology and software to build the virtual assistant, but departments such as customer service or human resources cannotwork properly with it. This is because they have fundamentally different technology needs. It is very important for IT to have the flexibility to choose the right technology stack for the virtual assistant and to be able to switch between NLP services without a technology lock-in. In addition, an open system with APIs is needed to integrate enterprise systems, analysis systems or other functions. Whereas business units like human resources, customer service or marketing departments, that will run the chatbot to get in touch with customers, have other needs. They have to take care of intent management, analyze data and need a handover tool which handover chats to real employees if necessary. A software that meets the needs of an IT department does not necessarily cover the needs of content / intent managers or the departments. If the different needs of those involved are not equally taken into account, problems will inevitably arise in the long term.

In summary, there must be a solution that meets both IT and departmental requirements, preferably combined in one product.

The third problem: the choice of technology

Choosing the right technology for Conversational AI projects is quite difficult. The market is full of different technology vendors and NLP services. It is difficult to keep track and judge which provider to choose or who will prevail in the future. Another challenge is that many vendors only offer a partial solution, such as natural language understanding and processing, a channel connector or an analytics system. In addition, it is often noticed during the course of the project that the selected technology does not meet the requirements of the company. For example, after some development time it is noticed that enterprise systems cannot be connected or that certain APIs are missing. As we have learned, there are many complex requirements and needs for companies to successfully scale and run virtual assistants.

Of course there are many more reasons why chatbots are not successfully operated. Often, important aspects such as a good user experience, which should be ensured from the very beginning of the conceptual design, fail. This can only be achieved with project participants who have experience in conversational design or copywriting. This and much more is shown in our whitepaper "3 Reasons Why Chatbots fail" and treated in more detail.