The 7 traps to avoid when creating a chatbot
Chatbots, if poorly constructed, can turn the acquisition of a potential customer into a bad user experience:
Some bots always give the same answers because their database does not allow them to have more vocabulary. Others fail to answer the simplest questions, because their creators did not take into account that users ask questions that are sometimes stupid, sometimes unpredictable and sometimes both at the same time. Worse, some chatbots misunderstand a question asked in their own area of expertise.
To prevent these failures, here are some pitfalls to avoid:
Try to solve all problems, for all types of users
It is impossible to create a chatbot that can resolve all customer requests. Cortana or Siri could possibly develop this capacity but the best is to create a bot specialized in a domain and with a precise goal. So he will be able to really engage your target.
How to do it ? Define the purpose of the chatbot and in which cases it will be used. Don’t try to solve problems that are outside your area of expertise, instead focus on mastering a specialty. Manage user expectations by trying to keep conversations in the bot’s comfort zone.
Do not work from a use case
Most chatbots still need to be human trained in order to generate the best engagement results. Deploying a chatbot without establishing a scenario upstream is nonsense. In addition to a good scenario, all the problems that a bot won’t be able to solve must lead to a qualified person who will answer them.
How to do ? You have to create the main scenarios, and use all the new data acquired to reinforce the knowledge of your chatbot. If you master this, your team will be able to spend less time responding to customers and more time dealing with important tasks.
Use too limited a linguistic style
This major defect can be defined as:
1. The inability to answer complex, personal, contextual, or unusually phrased questions.
2. The overuse of emoticons and “casual” expressions
3. The almost non-existent response to sarcasm, humor, insults or onomatopoeia.
How to do ? We must start by defining user expectations. If necessary, let your bot admit that it’s lost and put the conversation back within the limits of its understanding. Also reinforce the learning of your chatbot by improving its database. It will only be better in the next conversation.
Have a non-existent or ill-defined personality
Some users may enjoy talking to a robot that speaks like a robot. For others, they will need to feel like they’re talking to a real person. This is why chatbots should be as personified as possible.
Be careful, using random expressions, slang or emoticons does not work.
How to do it ? The tone, the vocabulary, the avatar of the chatbots are all necessary elements and to define to create a coherent personality. It must also be in line with your target. After all, the bot represents your brand as much as a human being …
Poor integration of the bot into business processes
A bot that has no consistency with the company it represents or with the type of user with whom it exchanges is very dangerous for the brand. Users don’t like to waste their time, and they can lose patience if a bot keeps asking them for account information or is unaware of the latest transactions.
How to do it ? The bot must be allowed to access each user’s information and update them in real time. You have to give him the contacts of key people in the company and all other information that will help him create a relevant value proposition while reducing the steps to hire each new user.
Neglecting security and protection of info
Remember, Twitter users turned Microsoft’s Tay bot into a misogynistic, neo-Nazi chatbot in just 2 days.
Since the chatbot represents your brand, such an action could be devastating, and more particularly if you integrate your bot into your CRM, your internet servers or any other sensitive system.
How to do it ? You have to adopt good security attitudes and check that the chatbot is effectively protected against external attacks.
Don’t think of “conversational UX”
The benefits of using chatbots are increasingly visible. Most businesses are going to want to use them to support and improve their customer service, sales, product development, and social media.
No matter which bot you want to develop, keep in mind that it is the user experience (conversational UX) above all that will make your chatbot successful or unsuccessful.
Thomas sabatier is the co-founder and CEO of The ChatBot Factory, an agency specializing in the development of conversational applications. He is also an associate director at BIGMITCH, a digital innovation consulting agency. He started his career in e-commerce, before joining a consulting firm as Key Account Manager.