Imagination Technology
Natural Language Processing and Dialogflow
“You never change something by fighting the existing reality, to change something, build a new model that makes the existing model obsolete” Buckminster Fuller
Combine imagination, technology and talent. Lets call it Imagination Technology. What use is information technology without a healthy dose of imagination and talent applied in creative ways? Without it you just end up with more of the same — same results, same pace of change, same excuses. Natural language capabilities are now sufficiently mature to apply to many business problems in a creative way.
Voice Assistants and Speech to Text
It’s quicker to talk to your phone than it is to type. At least 3 times faster. The shift from typing to speech will happen incrementally and in tandem with improvements in the interfaces. We are considering this in our development and how we support our customers, even though most of the team and our customers are still typing.
The massive growth in digital voice assistants such as Google Assistant, Alexa, Siri and Cortana and the coming wave of voice enabled hardware also warrants our attention. Future customers will expect to have an option to engage via these devices and we need to think of integrated solutions to replace certain archaic digital experiences. Will our customer of the future navigate to our site or portal to get the information they need? Or will they interact via speech and an associated hardware device to engage with one of our competitors?
Accurate speech to text and working voice assistants until recently were figments of our imagination. Using our imagination and the available AI technology is a great opportunity.
Natural Language Processing & Chatbots
We are working towards using Natural Language Processing (NLP) to achieve greater automation, operational efficiency and agility within our customer service functions.
Natural language processing is the ability for machines to interpret the written or spoken word, and is a key component of automated intelligent agents, also known as chatbots. Chatbots can interpret human interactions and respond autonomously. They can be built quickly, or you can partner with a provider, surfaced on most digital mediums such as Facebook messenger, websites, Skype, Alexa and are an efficient way to resolve basic customer queries in a cost effective way.
Death to the FAQ
At the next opportunity we hope to complement our support channels with new natural language solutions and an entry point for voice enabled interfaces. A baseline experience is that each simple query should be answered seamlessly, and be repeatable via an autonomous process each time it is asked. We need to provide a means to gather the questions customers ask and not assume we have all the answers, that are then built on a frequently asked question page.
One of our teams support a large network of financial advisers. At a high level the simple requests managed are system FAQs from both new and existing customers, prospects seeking information and wanting to be contacted and other searches for information. For complex requests, related to CRM data and specific customer transactions, these can also be managed by a chatbot, but were outside the scope of our simple build.
The Platform — Dialogflow (Previously API.AI)
Part of the rise in chatbots can be attributed to the open APIs available to developers. We used the Dialogflow platform (thanks Dawid Naude!), to build the NLP component. We then integrated it into a custom business page on Facebook Messenger.
The integration is easily configured and the same back end processing in Dialogflow can be surfaced on other communication platforms such as Skype, websites, Cortana and many other front end interfaces, as displayed in the Integrations image.
Beware of solutions that you can only surface on one channel, for example, licensing a solution that can only be used on Facebook messenger. Any solution should be easily scaled across many channels utilising the same NLP logic.
The Solution
We built a basic chatbot solution in under 2 hours that could answer many text based inbound requests. A basic Facebook business page for a fictitious firm called AI Financial was created and the chatbot integrated into messenger to manage simple requests. The 2 conversation examples used were capturing information from a prospect and answering FAQs. It was relatively straightforward to train the chatbot to handle exception scenarios and most of the work was done via configuration. When Facebook Messenger opens up to speech to text the chatbot will work in the same way.
We were most impressed with how the solution independently adopted its responses and conversational branches based on previous interactions. General questions were more quickly routed down specific paths based on past behaviors. This occurs via machine learning.
What We Learned
- Do not think of chatbots as a replacement for quality staff, they can complement a team but not replace the real human value.
- If you utilise a chatbot, clearly label it as a chatbot, as it will screw up sometimes. Don’t be limited by traditional communication channels and mediums to support your customers.
- Do not default to live chat as your next channel for customer servicing, think beyond this to a chatbot that can handoff to a person if required and or integrate into your CRM to utilise data.
We are really only limited by our imagination and our fear of change when it comes to using the latest technology to power elements of our business.
We are trying to tackle the challenge of competitiveness and customer engagement with the latest technology in mind. We are also trying to move beyond the bottlenecks of a traditional business model in order to adopt new technology and stay relevant to our evolving customer base.
The purpose here is to share simple practical AI solutions on accessible platforms that could be useful to test out your ideas. It’s not to delve into lots of technical detail and compare options. If you want to explore chatbots in more detail, this is a useful introduction.
Thanks to Lucy Adelaide for working on the illustrations above.
The views expressed in this article are those of mine alone. This is part 5 of a 6 part series exploring practical business applications of artificial intelligence.
