Building simple chatbots with a cloud platform is easy. But when you need chatbot technology for a company in the financial sector, things get more complex.
The first step is to understand the needs and constraints of your users.
In my case, I identified two main pain points. First, the privacy of client data was an absolute requirement. On premise was also a must-have. Financial institutions have multiple use cases, and they are complex. As just one example, consider a chatbot that manages the lifecycle of a mortgage. This involves lots of data, complex products and rules, and compliance issues. It is definitely different from the pizza-ordering or weather chatbots we are familiar with.
The second step is to shortlist the technologies.
There are dozens of online chatbot platforms, like Wit.ai from Facebook and Api.ai from Google. But these are cloud-only, so I was able to easily rule them out. Even beyond that criterion, they are best at building simple chatbots and are not suitable when you need to handle complex use cases comprising hundreds of variables.
Then there’s IBM Watson Conversation, the Watson API for chatbots. IBM has a long-established relationship with Fortune 500 clients, so the trust factor is already in place. It’s not on premise, but they have a premium dedicated cloud offer. It’s a serious pick.
ChatScript is an open-source solution with a fantastic history. ChatScript’s inventor, Bruce Wilcox, didn’t like the existing chatbot technologies so he decided to develop a new language from scratch, and he went on to win the Loebner prize four times. (The Loebner prize is a Turing test in which judges trap chatbots with tricky questions, and while it is quite controversial, it proves the idea that very complex chatbots with thousands of rules can be built with ChatScript.)
Amelia from IPSoft is presented as a “digital employee” and a “cognitive agent.” I had the chance to get a technical demo of Amelia, and it looks like a very good platform for building top-notch chatbots. The price has not been made public, but I don’t expect it to be cheap.
I finally decided to proceed with ChatScript. It perfectly met my constraints: It can be installed on premise, and I knew that I could build a complex chatbot with it. What’s more, the chatbots are authored through real source code in plain text files. This is key because it allows source versioning and collaborative authoring and it further triggers continuous integration and non-regression abilities. Finally, there’s no need to mention that I was excited to use the Loebner’s winning technology.
I was delighted with this choice — in fact, “How I went hands-on with ChatScript” might be my next article.
If you’re looking at chatbot technology for the financial sector, IBM Watson Conversation, IPSoft Amelia, and the Loebner’s open-source winner ChatScript are my top picks.
Ludan Stoecklé is the CTO of Addventa, an AI engine for business solutions.
The audio problem: Learn how new cloud-based API solutions are solving imperfect, frustrating audio in video conferences. Access here