Chatbots: a must, or a potential trap?
No CMO can present at a marketing-related conference without including a couple of slides about his or her company’s chatbot. Following the multi- or omnichannel strategy and customer journeys, it’s now the turn of the chatbots. And that’s only logical. Chatbots are the future, they’re innovative, and they show how progressive an organisation is. Don’t they?
The word tells you everything you need to know: it’s a robot that chats. That’s its primary function. A chatbot can help consumers with answers to questions about an organisation’s services and/or products. The purpose of each message is to solve a question, whether or not it’s linked to a transactional component.
Chatbots promise to reply in a relevant manner in real time. Companies or brands use this to increase their capacity to help current or potential customers at any time and anywhere in the world. Facebook, WhatsApp and a handful of other initiatives offer the platform and tools to give your company a kick-start too.
Chatbots are, without a doubt, technologically innovative. A chatbot uses machine learning to properly assess messages in a particular context and with a certain type of relevance. Over time, it gets better and better at doing so. For consumers as well, at first glance it appears that this is primarily advantageous. In principle, there is no possibility of incorrect interpretation, and the effect of non-verbal communication is completely excluded.
Social nuances and cognitive efforts are negated by chatbots. This is in line with our natural psychological preferences in the search for further simplification of our existence. As consumers, we want things to be made as easy as possible for us. It’s not without reason that there must always be a possibility of interacting with a real person during a chat. Natural Language Processing (AI) is still incomplete for as well the Dutch as English language, but has the potential to reduce the number of necessary human interventions.
A chatbot: an end in itself?
There are still great challenges. The main problem is that companies see having a chatbot as a goal in itself. In fact, a chatbot helps to efficiently improve the familiar customer experience. Here we go again: efficiency. It doesn’t say anything about the effectiveness and often also nothing about the long-term experience. In the short term, there may be a positive boost to the customer experience. At the same time, a chatbot is currently unable to replace a company’s own service staff.
At the moment, many chatbots lack the empathy that your own employees generally have. This is why a chatbot is not a standalone novelty, but something that is provided on top of the services already on offer. The development of a bot for your organisation is ‘relatively’ simple, but developing one that’s an extension of your brand or organisation and that takes away some of your customers’ questions is a story in itself. If you’re not able to directly fulfil your consumers’ expectations with a chatbot, based on their experience with your organisation through other online and offline channels, the question arises as to whether something innovative reflects your long-term goals.
Another challenge is the way the chatbot is seen within your organisation. As a means of reducing costs? No way! A further challenge is that you have to demonstrate to your own service staff the added value of bots. A chatbot is after all not a replacement for your own staff, but an extension of them. Not only will the chatbot need to be developed, but your staff will need to be trained.
Set a bandwidth
Especially in the beginning, a chatbot will need to be monitored, filled and adjusted. One of the most important matters is to determine the bandwidth of chatbots. In China, a number of chatbots such as Tay, BabyQ and XiaoBing, produced by companies like Turing Robot and Microsoft, have been taken temporarily offline after they responded very explicitly to specific subjects such as the Communist Party or the Chinese territorial waters. Xiaofeng Wang, a senior analyst at Forrester, said earlier to Financial Times: ‘Machine learning means that data will be used. In this case, that data is also available somewhere on the internet. If you don’t indicate a clear bandwidth, you can’t determine what is being learned.’
Measure your chatbot’s success
Marketers have become rather spoiled in recent years. In addition to reach and commitment, matters such as customer satisfaction and incremental sales relative to control groups are reported. The complete sales funnel can be monitored, and where necessary, further optimised and the ROI reported. In the area of messaging apps and chatbots in particular, there are no standard metrics available yet. This is therefore the next discussion. With respect to , matters such as average session time and confusion triggers can be considered.
A bot is ‘just’ a bot
It’s not just the chatbot itself, but also the complete process and set-up around it that will make the difference. Good technology does not yet make a good dialogue, but good dialogue supported by good technology makes a better dialogue. Start with the human factor and creativity, instead of with the technology.
This article is also posted on the Dutch platform Emerce.nl