People are talking a lot about Chatbots, and will soon be doing a chatting with one and the primary industry adopting this will be customer service. Chatbots will help people converse with computers in their native language via a computer interface. With the rise of messaging apps, the explosion of the app ecosystem, advancements in artificial intelligence (AI) and cognitive technologies, a fascination with conversational user interfaces and a wider reach of automation are driving the Chatbot trend. With the proper assessment of business rationale and implementation, there will be quick near-term results and longtime justification.
Chatbots will not masquerade as a human but will provide first and quick support – saving time and energy – for both the customer and service providers. These will heighten business outcomes and deliver superior experiences, and will continuously learn as they “chat” with humans, providing meaningful responses to any new queries and unique requirements over time.
An Understanding of Terminology Around Chatbot’s
Every person starts a chat with an intention. This could be anything like Check Order Status, Cancel Order, Return Item, Refund Status, etc. Based on an entry made by the user in the Chatbot, the engine will map that to a specific and discrete use case or unit of work. Above all are “use cases” that a Chatbot should support and from the free text entry that the customer types in, a chatbot will work out exactly which unit of work should be triggered in the background. Oracle Chatbots will break text entered by users into words and ranked against intents to respond with most accurate details.
Having defined an intent (a unit of work) it will always be derived based on common phrases that will be entered by the customer. You can think of this as a kind of sample data form which the AI/ML element of the Chatbot “learns” from. So, for each intent, you would define a pool of utterances which are used to help the intents engine work out which user input maps to which intent. The nature of the language with subtle differences means we can’t do a simple string match. Instead what happens is all these utterances are used to “teach” the intent engine. That way, subtle differences in the input such as “What is my refund?” vs “What’s my refund?” or “Tell me my refund amount.” will be mapped to the same intent.
In some cases, you might want to think of an intent as having a variable element – like a parameter. For example, “cancel order” will result in more than one action like cancellation, return, refund. Using entities, we can define that certain elements of the input should be identified as these variable values. Oracle Chatbot gives more flexibility and power with these entities as it will help us build multiple modules.
For a Chatbot to be more successful and useful to both business and the customer it must be data-driven rather than task oriented. The data could be coming from both a customer as well as the backend ERP systems. Oracle Chatbots are language independent and accept free format text and is backed by the powerful AI engine developed by Oracle.
Like any conversation a user will start their conversation with the Chatbot saying hi/hello or even with a question “What’s the status of my order?” The Bot will take the words from this utterance and try to match and rank with the intents defined. Based on the intent match the Bot will return some responses and as the customer continues the dialog flow commences. The dialog flow can take various directions based on user’s intent and their utterance.
Oracle Chatbot Implementation Architecture
The following diagram provides a high-level idea of various layers of Oracle’s Chatbot solution. The UI/user interface will be one or many chat channels, these can be any existing popular channels or bespoke channels or web applications. A user can initiate their dialog with a Chatbot from various channels wherever we configured it. Channels are configured through webhooks. In the backend, we have the message platform and underneath that, we have the Oracle homegrown AI bot engine. To back them, you have Oracle MCS backend where all your custom components can be built. You can reuse all existing MCS backend services which you have already built or build new ones.
As you can see, an integration layer is optional but it is strongly recommended so that you can separate your integration’s or microservices. Again, you can use all the services which have already been built and keep building the ones you don’t have. Using this integration layer, you can integrate the Oracle Chatbot to any SaaS, PaaS, On-Premise applications to execute your business services. This can start from basic order inquiry to processing cancellation/refund. You can keep all the complex business rules, validations, executions in your integration’s/microservices and let the Chatbot do the interaction with the end customer. We can make a seamless and pleasant experience for the customer.
In the current market, we have two types of Bot solutions Task-Oriented and Data-Driven & Predictive. This shows a high-level differentiation between the two. Oracle Chatbot falls under later one.
With a Chatbot, businesses can quickly develop and host simple to complex bots in very less time. Oracle chatbots are language independent and it allows you to host your bot in various languages keeping the core dialog flow and integration’s the same. Having a Bot will help companies to provide faster and better service to their customers. Customers will have the option of chatting with a bot and getting the info or service immediately rather than calling customer service desk and then kept on hold for minutes before they get to an executive.