Oracle weighs in on chatbots for enterprise developers

We recently had a conversation with Suhas Uliyar, Vice President, AI, Bots & Mobile Product Management, Oracle to get some insight on chatbot tech in general, and how developers in the enterprise can leverage the use of chatbots.

ADM: How have you seen chatbot technology, and adoption of it, change in the last several years?

Uliyar: Over the last decade or so we have seen the adoption of mobile as a primary engagement channel for consumers and enterprise employees alike. What we are seeing now is an increase in messaging channels like FB Messenger, WhatsApp, WeChat, Slack and SMS for all conversations. There are billions of people around the world on these apps and users want the same experience when communicating with businesses as they do with friends and family. Over 4.1 billion users around the world are on instant messaging and chat apps at any moment in time. People want and expect the instant reaction that only messaging apps can provide, and they’re rushing to these platforms in droves, at an adoption rate far greater than even social networks saw in their heyday.

The rate of adoption by enterprises has shifted dramatically. This time last year, many thought it was a good concept and customers were intrigued, but now we are seeing customers assign budgets and piloting programs at an incredible rate. We are seeing customers that missed the mobile stage jumping straight to the chatbot world. There are a few factors contributing to this increase, including the strong ROI chatbots offer, the fact they represent a strong use case for adopting AI in the enterprise, and the minimal budget required to deploy across multiple channels.

ADM: Do you think the growth of consumer devices with voice capabilities have normalized bot interactions?

Uliyar: It’s amazing to think the iPhone just turned 10 years old. It was one of the first to introduce a voice assistant with Siri, and new voice assistants have added powerful technology and a variety of integrations that can make them very useful. Adoption of these is slowly increasing, but they are laying the groundwork for making the technology an accepted part of our everyday lives.

In the meantime, mobile has taken over the world; we’re half way towards everyone in the world having a smart phone, and are used to having an app for everything. Now there are so many we are getting app-fatigue, and only use a small handful of the ones we download for any significant time.

Instead, we are turning to messaging services, which are becoming increasingly prevalent, such as Facebook Messenger, WhatsApp, WeChat, Slack, Snapchat and the like, as well as digital assistants like Google Home and Alexa. They satisfy our demand for immediacy of response, and being conversational, give us a greater level of ease when interacting with them.

ADM: What benefits do chatbots offer enterprises and end users over traditional call centers, FAQs, etc.?

Uliyar: Traditional call centers can be limited without the expansive data and intelligence behind modern chatbots. Using machine learning and natural language understanding (NLU), bots can apply sentiment analysis, reference required data or applications quickly, and resolve most consumer concerns without human intervention. This frees up time for customer service employees to focus on more critical situations, whether it is taking over a conversation with a frustrated customer, completing a process a bot cannot address, handling priority customers, etc. Chatbots rapidly deliver engaging conversational experiences backed by enterprise data across numerous systems and applications.

In addition, bot interfaces allow customers to state their concerns and immediately begin resolving the situation, rather than going through a phone-based decision tree and hoping they get transferred to the right department.

ADM: What are businesses looking for in chatbot services and how has Oracle developed its mobile platform to meet these needs?

Uliyar: Companies are empowering their employees with chatbots alongside mobile apps for mobile field service, sales, employee self-service and expenses. They are looking for services that can easily deploy and integrate with the systems already in place, streamline processes, and provide analytical insights to gauge effectiveness.

Chatbots are taking off now for three big reasons. The first is that with messaging channels, we tend to get an immediate response. The second is that we can engage with bots the way we naturally talk. And third is that several technologies have matured to where they’re ready for large-scale, customer-facing duty. Those technologies include natural language processing (NLP), artificial intelligence (AI), access to large sets of data and cloud computing. And lastly, cloud-based IT platforms provide access to high-performance computing and can make bot building easier. Using a development platform, a company can quickly build and deploy chatbots using drag-and-drop screens, without having to hire specialists in natural language and machine learning technologies.

We have continued to innovate and have expanded Oracle Mobile Cloud with chatbot functionality powered by AI that uses machine learning algorithms to help enterprises engage with their customers and employees through chatbots. There are several machine learning algorithms in our solution to process natural language from the users, deep learning algorithms for the chatbot to continue evolving, and algorithms to understand user sentiment or language. The solution we have built has four major parts: 1) to integrate with these various channels, 2) have a wizard to model the dialog with the end user 3) AI / ML algorithms and 4) integration to enterprise data.

Oracle Mobile Cloud has been adopted across several industries including retail, financial services, industrial manufacturing, automotive and construction. We have several customers with business-to-consumer use cases in the banking, travel, retail, utilities and hospitality sector for both transactional as well as service related conversations. In parallel, we are seeing use cases for HR like employee on-boarding, talent acquisition, and sales CRM, approvals, time cards, and leave management for ERP.

ADM: Why do you think it’s important to make bot implementation easier for developers?

Uliyar: Our goal has always been to make it really simple. Not all developers want to get in the weeds, so we have tried to streamline the process for those who would rather be focusing on other parts of the business. Bot technology brings great potential for businesses, and it’s important to make it as accessible as possible for back-end users and end users. With the introduction of Oracle’s intelligent bots, Oracle will be able to embed chatbot capabilities into its portfolio of cloud services, allowing developers to focus on tuning the chatbots to meet line of business use cases - for ERP and HCM users or to meet industry specific needs of retail, telecommunications, and others. With Oracle’s intelligent bots and Bot Builder, developers get out-of-the-box capabilities that allow them to easily build a bot for enterprise employees or customers, horizontal use cases and industry specific domains with personalize engagement experiences for the user.

With Oracle’s offering, developers get a multi-channel platform, linking user experiences across bots, mobile and web. It also provides an integrated solution that brings together channels, dialog flow, AI engine, and integration with bot builder UI. Organizations can now better engage with customers and employee across all of today’s most popular messaging, mobile and web platforms. We have introduced a number of features to support this effort, including the ability to abstract channel interfaces, fine tuning and optimizing of NLU, promising a continued evolution through additional algorithms and enterprise integration capabilities. Of course, delivery via cloud also reduces the cost, effort and complexity of managing patching, and provides a reliable and secure cloud that handles backups, high availability, fail over.

By enhancing and optimizing the NLU, which is essential to understand the conversational interaction with the end user, we can provide a pipeline of algorithms based on linguistic modeling, deep learning neural nets, and spectral word embedding, and we automatically pick the best model based on the customer’s data set (or lack thereof). Additionally, a platform that automatically extracts pertinent information from the user’s conversation (such as date, currency, time, etc.) and slots it for integration to the backend systems of record eases the implementation process for the developer.

ADM: What features or capabilities have developers been asking for? What has been their biggest challenge up to now?

Uliyar: Developers building chatbots into applications must start by thinking about the type of interactions the business wants to have with users - for example is it sales requests, customer service inquires, or information requests? Once these are established, a developer needs to look at ways to increase the accuracy of processing natural languages from the end users with the hope of more effectively and efficiently providing support and resolving customer issues. NLP is now good enough for bots to understand everyday language and give conversational responses. AI and machine learning now allow bots to continuously learn the different ways people ask questions and to let bots respond with personalized answers.

They have also been looking for ways to easily integrate the most popular messaging clients including Facebook Messenger, Slack, Kik, and Line as well as digital voice assistants such as Amazon Alexa and Apple Siri. This isn’t a simple task - each has different nuances around how it handles queuing, error rates and so on. Basically, a lot of things happen in this integration layer, so without going in to gory detail, this is something we have taken out of the equation. We now provide a declarative UI that manages all of that.

ADM: What are ideal use cases for chatbot-powered applications?

Uliyar: There are many uses from adding personalized services and improving the customer experience and reducing the cost of customer service. We are seeing customer interest and use cases across every industry – from retail and banking to utilities and ecommerce. As an example, we have a company in the utilities industry that is using chatbots to help customers check and pay bills, see their energy usage levels, answer the most commonly asked questions and report or find out details about outages.

Consumer banking is a prime use case for chatbots. Normally in online banking you check things like your transactions, your balance, check on payments or transfer money on the bank’s website or on your phone. Now with chatbots you can use NLP to ‘talk’ to the bot through a messaging service or Private Virtual Assistant (PVA) in a conversational manner. Users can ask questions like ‘What is my bank balance?’ and to undertake activities like transferring money or paying for something.

ADM: How will machine learning and AI applications impact chatbots and the user experience overall? What will this mean for human customer service workers in the future?

Uliyar: We believe there is more to chatbots than simply adding AI. Our chatbots platform is built on a micro-services architecture with machine learning engines designed and operating as a micro-service. For the chatbots service, our channel framework integrates with Facebook Messenger, WeChat, WhatsApp, native mobile apps, web-based apps etc., and as these conversations come through the platform, the NLP micro-service handles the language recognition aspect. This same NLP service can be invoked by any developer that wants to build an application that utilizes NLP. This architecture allows developers to use specific machine learning algorithms for their use cases such as using sentiment analysis to change offers to a customer or using a recommendation engine within their consumer app.

One of the keys for bots is to also handle complicated information flows. If I want to pay my babysitter, I might need to tell the bot to pay them from a set account, or I might want to check if I have enough cash in that account first. Having confirmed that detail you don’t want to have to tell it which account again. Instead you want it to take the flow from conversation and realize you are still talking about the same account. So as you can see, context is key to providing a good experience, as typically humans don’t think linearly.

Chatbots are great for conversational UI that involves using unstructured data to interact - it provides a great experience where users are performing transactions or looking for answers to questions - but the conversational UI is not the ideal interface for capturing a large amount of structured data or displaying a lot of data. An app or a form is the best way to display this but this needs to be done without losing context or requiring the end user move to a different channel and so we expect to see greater use of mobile instant apps, like Chatbox, by bot applications.

Getting the NLP and AI elements right is very important. It has to be able to understand the questions asked even if they are using colloquial terms or put across in different ways. So for the banking example I just mentioned, you could ask the bot for the balance in an account, or you might ask it “Am I broke?’ It has to be able to understand the slang.

ADM: How do you expect bot technology to evolve in the next 5-10 years?

Uliyar: We are just at the start of conversational AI as intelligent chatbots. While these services need to be authentic and not try to pretend to be humans, we are continuously adding additional algorithms around user sentiment, image analysis, language translation, self-learning and behavioral analysis, to both simplify bot development and to enhance the user experience.

We believe bots will be everywhere and conversational UI will be the primary interface of interaction. There will be a convergence of bots and immersive UI (augmented reality and virtual reality) with devices that are yet to be imagined but will be used via a conversational UI that interacts with a bot.

About Suhas Uliyar

Mr. Suhas Uliyar is a 20-year+ mobile industry veteran, known as a visionary, strategist and technology evangelist responsible for designing and developing enterprise mobile & intelligent chatbots, platform and tools. Suhas was named in the Top 100 Wireless Technology Experts by Wireless World in 2014. Suhas is responsible for driving Oracle’s mobile strategy and vision. He is a seasoned executive with years of technical and business management experience in enterprise software, artificial intelligence and machine learning with a successful track record as both an entrepreneur in small start-ups and an executive with major industry leaders. Suhas has held leadership positions with SAP, Motorola Solutions, Spring Wireless, Dexterra (Antenna Software) and Micromuse (IBM).