Building bots on a serverless architecture
At Kitsune we decided to merge a rather challenging web app with a seamless platform and built a function that will take enterprise marketing to the next level.
Having a chatbot integrated in your website for either marketing or customer support adds to your acquisition and retention strategy.
24x7 human customer service is indeed, irreplaceable. However, we cannot neglect the productive automation solutions that come with lead capturing, raising grievance tickets and then integrating chat interfaces with atlassian or equivalents like confluence to follow up with the processes.
Yes, you need a chatbot but why does it have to come as a web app built on serverless technology? Scroll down to know -
The Anatomy of a Chatbot -
All chatbots come with some basic components- NLP, a dialog manager, content and prompt events. The crux is that a chat interface is the user’s window to trigger some real-time events in expectation of a real-time response.
1. Interaction layer - The IM, Chat window or Social media interface for human interaction.
2. The connecting orchestrator - To ensure incoming and outgoing of data.
3. The NLP core (AI begins here) - The NLP core is where the AI magic begins, the sentence to vector conversions with cleaning of stop words, punctuations and demarking of key phrases happens here.
4. The Machine learning block - The ML block is where the processing of data happens using a RNN -(Recurrent Neural Network). Or for a decision tree based (or prompt based) chatbots t. In either case, it’s regression or classification, or both, that makeup AI.
5. The Business logic layer - (Narrow Artificial Intelligence) is what helps return specific communication for a very specific use case such as banking, risk management.
6. The API point layer -
The API point layer is how you connect your chatbot to a platform/framework. This platform/framework then manages the loading, databases and virtual machines.
Chatbots may be based on simple decision trees or they may come with a complex RNN (Recurrent Neural Network). In which case there will be an LSTM(Long short-term memory) - basically a set of mathematical algorithms that allows machines to retain points to build a sequence that the RNN can retrace in cycles, to replicate recollection in human neural networks. This is deep learning and this separates the more advanced chatbots from a decision tree based - machine learning bot.
Irrespective of the underlying technology, Machine Learning or Deep Learning, chatbots need huge memory databases and processing. No prizes guessing that like any application that has such ‘great expectations’ of the space-time complexity, chatbots will work best on a serverless architecture with loading optimization and high availability of databases for a seamless user experience.
Building A Chatbot Application with Kitsune -
API layer choices —The 3 most important options while choosing API partners for chatbots are the -
1. Availability of the API platform
A serverless architecture implementing the reactive systems will ensure an elastically scalable, responsive, message-driven service, agnostic to technology and backend.
Kitsune is built to align with the reactive manifesto which is ‘jargon’ for a well managed distributed system design. In the case of a serverless technology - over the cloud. Thus building any web application that requires delocalization and efficiency in terms of robustness and speed needs a platform like Kitsune.
At an enterprise level when you have multiple locations, to function out of and multiple domains to cater to in terms of business and customers, you would not want to be worried about the backend or the NLP core or the AI block. We also give you the ANA API atop Kitsune to ensure that all you have to do is build a beautiful interaction layer for your end user.
Take care of the chats while we take care of the noise!