More and more organizations have placed a chatbot at the bottom right of the homepage of their website. As soon as you visit the website, Nina, Billie, Nikki or Sam will introduce themselves to you. Asking questions is usually not a problem. But a good answer? In this article, we provide the seven pitfalls for organizations that want to add a chatbot to their customer contact channels. 

 

1. “We now also have a chatbot” – or is it an interactive Q&A?

Placing a chatbot on the homepage invites a dialogue. A dialogue consists of a free question and answer game. If necessary, other information should be consulted in such a dialogue in addition to the customer's input. Not every bot has that option. When your chatbot works on the basis of a tree structure or flowchart, it is more likely to be an interactive FAQ or Q&A. It is not for nothing that these FAQs are usually on the contact page. There are quite a few providers in the market that promise that your chatbot will function within a day, a week or a month. In that case, assume that you are dealing with a 'rules based chatbot' where your customer ends up in an interactive but limited system.

2. “Our chatbot is for 'simple questions'” – but does your customer know what those questions are?

Contact center managers are firmly convinced that the chatbot is especially good for handling simple questions. However, the customer has no idea what your organization considers a 'simple question'. There is therefore a good chance that the customer will talk to a chatbot, even if that is not wise given the nature of the question.

In addition, an answer can also raise a new question. Chatbot experts then advise to steer the dialogue in such a way that a new intent analysis is started.

3. Your customer doesn't ask for it and/or not all your customers ask for it

Are chatbots popular? Is the customer waiting for it? In the Netherlands, the reactions are rather mixed, but the chatbot is in any case not seen as the best invention of the 21^e century. The Consumers' Association surveyed 10.000 panel members, two thirds of whom are not satisfied. The chatbot does not answer the question (52%) or only partially answers the question (26%). Only 15% of the respondents find chatting with a robot no problem. Also in the online community of the Consumer Association the necessary feedback can be found.

4. You don't have the escalation option in order

“I can't help you with this” or “I don't understand your question” – that obviously doesn't help and is certainly not a first time fix. A 'dead end bot' costs the customer time, but delivers a customer experience that is comparable to that of an unskilled customer contact employee. Having to wait for an employee (because everyone is busy) who can take over the conversation after a bot has crashed may not be convenient. According to chatbot experts, there must also be a warm transfer during the transfer from bot to human. Then the employee can also bend the failure of the chatbot a bit.

5. “Our chatbot was already online after three months”

Getting started with a chatbot is quite a challenge. Among others one considerable period of preparation. In addition, a chatbot needs several optimization steps at the start. Those who focus too much on automating employee tasks and/or eliminating simple customer questions may not pay enough attention to improving the customer experience. What if it turns out that the quality of your service deteriorates with the introduction of a chatbot? Some companies are seeing a drop in the number of live contacts; very few companies see an increase in NPS.

6. You don't have a central knowledge base

Employees, FAQs, voice and chatbots, e-learning tools, solutions for agent assist or next best action, trainers and coaches: they should all draw from a single 'single source of truth'. Conversely, failed and escalated bot dialogues also provide food for thought for the knowledge managers. A chatbot needs to be maintained and that takes a lot of time if you don't also have your knowledge management in order. When information, products, prices, processes or procedures change, the designed chatbot dialogues must also be adjusted. In addition to a knowledge base, you also need a knowledge manager.

7. Take (future) legislation into account

If your chatbot runs on the basis of AI, you have to take more and more into account regulation from the European Union. The regulations surrounding AI are based on risk assessments. For example, the European Commission has classified the AI ​​of chatbot solutions as applications with a relatively low risk. But as soon as your chatbot also starts using algorithms that determine whether or not someone gets insurance, the chatbot solution falls into a higher risk category and more rules and restrictions apply. In other words: applications that still seem 'problem-free' at the moment may not be possible in the future.

(Ziptone editors)

Featured, Knowledge base, Technology