I was speaking at a meetup last week about a chatbot we developed for a city in Flanders, Belgium. Someone in the crowd asked how much personalization goes into the answers the chatbot gives. He asked if it would be possible to train the chatbot to take into account the history of previous conversations with the user to hyper-personalize the answers given by the chatbot.
An example:
- User: Where can I play football?
The chatbot could now either return a list of football clubs (organisations) in and around the city OR provide the user with an overview of available football fields (locations) in the area. Imagine the chatbot returns a list of football clubs, but the user follows-up on his first question by saying:
- User: I’m actually looking for a list of available football fields.
At this point, the chatbot knows his first answer wasn’t exactly what the user was looking for. The next time the user asks his original question (“Where can I play football?”), the bot could remember that the user was actually looking for a list of football fields rather than an overview of football clubs in the area.
However, the main point I want to make in this blogpost is that the intelligence is not in the bot. The bot is the interface. The intelligence is in the integrations and webservices that he chatbot can use. For instance, in the example given, this would only be possible by hooking up our chatbot to a recommendation or personalization engine. It’s not the chatbot itself that would ‘remember’.
There’s a common misconception that a chatbot = AI. That’s not always the case. A good chatbot will often use AI, for instance for NLP (Natural Language Processing). But not every chatbot requires AI capabilities. The main purpose of a chatbot should be to provide the user with relevant information or trigger an action in response to a simple query. And if that goal can be reached without any AI, that’s perfectly fine! Not all chatbots should have a tremendous vocabulary or be great at telling jokes, some use-cases simply don’t require those advanced capabilities.
So, what are some of those integrations you can think about when putting together your chatbot use-case?
- Incident Management, Case Management or Ticketing systems like ServiceNow, Jira, Topdesk or Planon. Most of these systems have an API that you can use to create tickets and allocate them to the appropriate customer support agent. The chatbot can be used as an interface to collect the necessary information in a more natural way for the user. The chatbot can guide the user through the process of correctly logging the ticket, and even offer a possible solution in case the problem is well-known. You can also use these systems the other way around, let’s say if you want your users to be able to check the status of the ticket they logged earlier.
- All sorts of calculation engines that can feed information back into the conversation. Let’s say you’re developing a car-insurance-chatbot that offers your users the possibility to calculate what a car-insurance would cost at your company. With the right integration that’s relatively simple, and the user will most likely prefer this experience over a simple static webform.
- CRM or sales applications. Take the example above; If you integrate with a CRM or other sales-tool, the data collected about your chatbot’s users could easily be sent to your sales reps so they can follow up on the most promising leads for your company.
- Integrating your bot with an RPA framework could automate a lot of manual, repetitive and time-consuming work. Think about generating temporary passwords for instance. Or copy-pasting information gathered by the bot into legacy systems that don’t have an API the bot can talk to directly. We’ve done a cool demo about this a few months ago with the team at RoboRana, that article can be read here.
There’s many more integrations we can think about: payment services, authentication services, the blockchain, your IoT network, … Possibilities are endless!
Thanks for reading! If this article triggered your interest or left you with some unanswered questions, feel free to get in touch.
This blog was originally posted here.