Hands-on with Chatbots (part 1)
I spent two days last week with a couple of colleagues in Oracle’s city office looking at Chatbots – or ODA (Oracle Digital Assistants) to give them their correct name. It was a really useful workshop.
I had been under the misapprehension that chatbots were earlier on the hype cycle than they actually are. Yes, a chat interface will replace most self-service UIs, but that’s 3-5 years off, right? Maybe not.
The growth of messaging apps over the last few years has been incredible. With 6 of the top 10 downloaded apps on the App Store being messaging apps and the spread of chat beyond just the workplace to groups of friends and families – spanning all generations – it’s a technology that almost everyone is comfortable with.
The workshop started with some case studies of companies who already have chatbots live. Examples included:
- The State of Geneva chatbot allowing citizens to access election results
- Hermes chatbot allowing users to track parcels
- Danish Technical University helpdesk chatbot provides FAQs, helpdesk tickets and passwords resets, plus a finance chatbot for expenses, procurement and invoices (integrated with EBS)
- Pernod Ricard chatbot integrated with JD Edwards
- Engie (2nd biggest utility company in the World) has a chatbot for Taleo to speed up the hiring process
- plus about 8-10 more covering government, gaming, sports, brewing etc
So it’s pretty clear that there are many examples of chatbots live, but were they working? It seems like the answer was ‘yes’ there also. Some of the examples above also gave success metrics. The Hermes chatbot has achieved a 50% deflection of customer enquiries away from service agents. They receive 17,000 calls a day and even when a chatbot cannot completely resolve the issue and needs to pass on to a service agent the human interaction time is reduced as the chatbot has collected much of the information needed. The payback time for the chatbot was an astonishing 3 days.
So, after convincing us of the viability of chatbots in the wild already, how do we go about creating them? I’ll address that in part 2 of this post.
Photo credit: Grant Ronald
Fusion Tipster 