When chatbots go bad
Chatbots offer the promise of great things for the future – but they also contain massive potential pitfalls and failures if they are not properly considered, designed and developed. There have been some recent examples of ventures from major international corporations where chatbot projects have gone badly wrong and damaged the reputation and business of the organisation. If you are going to get involved in chatbots you have to be aware of their limitations and conscious of the likely problems and difficulties in constructing something that works for both your business and the user and enhances rather than damages the user experience.
Here are some of the potential pitfalls and what you can do to minimise their likelihood.
- Limited vocabulary leading to repetitive or nonsensical responses: This is very likely to happen if you equip your bot with a restricted dictionary.There are several ways to address this;- Profile your likely users and the issues they will be pursuing and this should help you define and decide the language and phrases you need to include.
– Anticipate the likely questions and desirable, useful responses and program your bot accordingly.
– Make sure you cover as wide a range of topics and answers as possible so the bot doesn’t get stuck or just start repeating itself.
– Program the bot with alternative forms of question or issue so they don’t get stumped if something is put to them in an unexpected manner.
- Lack of integration with business model: If you fail to integrate your bot with your business it will very likely fail in its primary purpose. The bot is there to be a frontend interface in most cases and not a standalone function so it needs to mesh with your business systems, data and customer needs and expectation.For example, a user won’t expect to have to provide lots of personal data once an initial ID has been established. They will expect that information to be available to the bot so you need to link to customer data, purchasing history, search data and so on. It can also be important to have an escalation protocol so, if the chatbot does become stumped it can refer the user on and not just lose the contact.
- Lax security and privacy protocols: there have been recent cases of public-facing bots being corrupted and manipulated to a degree where they severely risked harming the company’s reputation, if not downright illegality. Chatbots need to be protected from this sort of external influence or you risk completely losing control of the bots possible responses and interactions. You need to restrict your bots ability to absorb undesirable words and concepts so they cannot reflect them back to unsuspecting customers!
As always, the only sure fire way to identify and rectify issues related to user experience in this area is by research and testing. The more you utilise research in both design and development, the more likely you are to finish with an acceptable, glitch-free interface and a chatbot that improves your service and reputation.
These are just a few of the potential problems (and some remedies) when dealing with chatbot development. If you would like to know more about chatbots why not ring us on +44(0)800 024624 or email us at firstname.lastname@example.org.