A large proportion of the chatbots that businesses deploy prove unpopular or ineffective. There are many reasons this can happen, but a few account for most of the failures.
- Poorly defined or overly ambitious goals. A bot needs to carry out a clearly defined task, and it needs to do it better than the existing alternatives. If users find that consulting the help pages gets better results for less effort than using a chatbot, they'll keep using the help pages.
- Failing to delimit the domain. The chatbot that can answer any and all questions isn't in sight yet. If it tries to cover too much ground, it will give unsatisfactory answers or say too often that it doesn't understand. The bot's introductory material should establish what areas it can cover.
- Not matching the bot to the audience. Style and tone matter almost as much as content. If the responses seem unnatural, condescending, pedantic, or flippant, the audience won't want to use it. The right tone for one audience could be entirely wrong for another.
- Poor development skills or rushed development. Creating a natural-language chatbot is difficult. Developers need to be specialists in the area, and they need enough time to create a good product. Trying to create a chatbot too cheaply will lead to trouble.
- Insufficient testing. A chatbot needs testing with users who are representative of the target audience. Developers need to take their feedback into account and make necessary revisions. A bot which is launched without adequate testing will get many user complaints.
Building customer satisfaction
The path to a successful product starts with a clear plan. It should define the subject matter, the expectations for its performance, and the business goals which it is expected to achieve. The task needs to be assigned to a skilled development team, which will deliver on a realistic schedule.
The final phase before release is testing the product with users who are similar to the intended audience. The bot should be released to the public only when testers report a satisfactory experience. After that, it's important to monitor usage to see whether the bot is meeting its goals.
