Authored by Suhas Uliyar, Oracle Vice President – AI, Bots & Mobile
The meteoric rise of chatbots, and proof in the last year that this rise is not a fad, has given end users a natural way to engage with business via a conversational user interface and has given brands new purpose in providing better customer service. As the brand-customer relationship has grown and has become even more intertwined, enterprises are beginning to see success with the deployment of these bots.
For example, Bank of America released Erica – a chatbot for consumer banking in early March, 2018 that had an adoption of 1 million users in the first 3 months of going live. Mutua Madrid Open, an Oracle customer, became the first ATP World Tour Masters 1000 and Premier WTA tournament to incorporate an AI-equipped chatbot to improve communication with tennis fans. Implemented with Oracle Cloud Platform, the chatbot, named “MatchBot,” used AI to maintain natural conversations that provided fans with information on the event, players, and results, as well as details on hospitality services, discounts on merchandise, ticket sales, access, and parking.
The University of Adelaide, another Oracle customer, created a chatbot on Oracle’s cloud platform to ease students’ pain during the university application process. On just the first day the bot was live, prospective students conducted an estimated 2,100 unique conversations with the chatbot, which led to a 40 percent reduction in calls to the University’s customer service line – and more impressively, a 47 percent drop in calls during the critical first three hours. In turn, this reduced the wait time for queries made via telephone – from an average of 40 minutes down to about 90 seconds, and 60% of student users rated their experience as “awesome.”.
These are just several such examples of successful business implementations using first generation, AI-powered, conversational interfaces.
So what’s next?
The next generation of conversational interaction is Conversational Intelligence. Today’s conversational interactions are either user-initiated conversations or notifications about pre-programed actions (e.g. Alexa notifying you that your order has been delivered or Siri waking you up in the morning because you have pre-set an alarm.) Conversational Intelligence, however, is the ability for an interface to “know” a user, learn and understand their moments, actions, behaviors and preferences, and recommend, predict or act on behalf of the user – essentially functioning as your own digital assistant. To put this in more simple terms, let’s draw an analogy to part of my job, which includes traveling to meet customers.
In my role as VP for Product Management, I have the pleasure of meeting our global customers and partners, and that's by far the best part of my travel experience. The reason, of course, is that the logistics associated with business travel can be difficult and time-consuming so any help is good. When my admin started several years ago, she didn’t know any of my travel preferences and I often had to initiate a conversation with her to book my travel. She would ask me my preferred airline, hotel, check on any personal commitments I had, schedule prep calls with local sales teams, make sure I had the correct customer briefing documents, and then would book my travel. On the day I left, I would ask her to book Lyft to the airport and so on. These conversations occurred a few times, but soon enough, she knew what to look for in my emails, which schedules to check (personal and business) and which flights and hotels to book based on my preferences, leaving me only to confirm her choices. As more time passed she did not even need me to confirm her selections as they were pretty perfect.
In this example, my admin learned to take actions on my behalf by observing my traits, behaviors, and habits, which she learned first by talking to me, and then by incorporating both her knowledge of our conversations and my validations of her actions into an overall understanding of my preferences that was accurate enough for both of us to trust each other when making travel decisions. Today, end users are having similar conversations with bots that may seem simple, but these conversational interactions are creating important data. Over time the incorporated AI built into these bots will learn from these data-heavy interactions, from aggregated data across, and from outside of, the enterprise will use this accumulated “knowledge” to become a functioning “digital” assistant. This is the future of where AI-powered chatbots are headed; a Digital Assistant for every consumer and every employee of a company.
What is needed to have a Digital Assistant for every consumer / employee?
- Naturally Conversational: Language is incredibly complex and the assistant needs to understand user goals/intentions in every part of the conversation – real meaning, contextual, personalized & human like.
- Knowledge, Memory & Reasoning – Users expect the assistant to behave like humans. The assistant’s "brain" needs to have semantic understanding of various knowledge domains, needs to understand events, needs to have long/short term memory, and needs to be able to make decisions and execute/orchestrate actions.
- The Assistant is for THE User – It needs to know users really well, their roles, behaviors, traits, preferences, their interactions and activities across various apps/channels. The assistant also needs to know how and when it should communicate with the user, and recommend actions based on their real-time needs.
- Proactive – The assistant should watch out for things the user cares about and either take actions autonomously or communicate with the user. Oracle's platform will provide capabilities to enable an assistant’s skills to be truly intelligent and proactive.
- Act as True Agents– The assistant should know how to get things done with minimal help – e.g. create a plan dynamically based on current situation/context, reason with data to decide and execute actions.
- Multi-Channel – The assistant needs to work seamlessly across various channels. This includes the ability to identify the same user in multiple channels, maintain state across channels and have a conversation with a user using multiple channels at the same time (e.g. for voice use cases where the medium itself is limited/constrained.)
- Digital Assistant Platform for Skills Development – Skill developers are a critical part of the ecosystem to make the assistant successful. We need developers to be able to develop skills that are intelligent, can detect complex situations, retrieve context from the assistant and be proactive in nature.
In my next blog, we will explore in detail how the Oracle Digital Assistant delivers on these requirements. First however, please go and listen to my podcast on the topic of Digital Assistants, which will be released on Wednesday, October 10th, here. It's short, easily digestible, and gives further context on Digital Assistants and where AI is going in the Enterprise.
