Kathy Gibson is at SAP TechEd in Barcelona – The humble chatbot can have a much bigger role to play in business operations as it gets smarter and more integrated.
Omer Biran, head of conversational artificial intelligence (AI) at SAP, points out that chatbots – or conversational AI – is about allowing users to interact with services and products in a natural way that is comfortable for them.
There are two user bases for conversational AI, he says, customers and employees.
The customer experience is pretty clear and there are many examples of how the technology can be put to work.
For the employee, chatbots can be used to reduce the complexity of systems, letting them interact more easily.
Biran explains that most employees are exposed to a variety of different systems in the course of their work. These systems can be quite complex.
“But users want to interact with systems in a simple way. They don’t want to have to run into productivity issues.”
Conversational AI can help with this, he adds. “Users can say what it is they want to do; then the chatbot can guide them through the process, or place them in the right place to do it.
“It is about making sure the employee experience is boosted and that they are more productive.”
Technologies like chatbots could also help to boost a company’s reputation with employees, Biran adds. “All companies are looking to recruit young talent – but if you have a has-been image where employees have to struggle with heavy SAP systems, this is not a winner.
“The old ERP interface was the old way of communicating with the enterprise, where the user had to master the system. Now, we are seeing the intelligence from the user to the solution – now I want the system to know me.”
Biran says that SAP provides a complete platform to build powerful and customisable chatbots, from end-to-end, and including training, connection to back-office systems, connection to front-end systems and more.
Users are accustomed to easy-to-use apps, so navigation needs to be a no-brainer in the enterprise, Biran adds. “Users shouldn’t need to master the structure.”
Creating a bot to help users with navigation is relatively straightforward, he says. Generally, the user’s intent can be deduced from their first sentence and getting to the place they need is simply a matter of choosing from a range of options.
A conversation is a bit more tricky and it’s not a straight line, but could take many turns or alternatives.
Future iterations of the platform will move to bot generation, he adds. “Today we are building bots: tomorrow we want to be in a position to generate the bots for customers. This would be a system that accesses all data services, and generates the processes and data services.
Another goal for future generations is the multi-bot, which will help users to navigate a number of applications. “Users will be exposed to many bots, in both SAP and non-SAP applications. Especially as a casual user, you don’t want to know what the different systems are and when you need to use them.
“The multi-bot will allow you to have a single point of entry, and will know which systems to engage with behind the scenes.
The challenges here are around mastering the user context. “We want a certain level of smoothness in the experience,” Biran says.
Application context is important too. “If I say something that relates to data displayed on the screen, I want the bot to understand this implicitly.”
The long-term goal is to make the chatbot the primary user interface, fetching data or applications as required and presenting them to the user within the context of the chatbot.
“This sounds like nothing much, but it could transform applications,” says Biran. “The chatbot could effectively become your digital assistant – and it is completely personalised because it knows who you are.
“We need to get to where the chatbot will read the context of the user and then it can feed the relevant data.”
The last focus point, collaboration and extensibility, is also evolving.
Collaboration will enable the chatbots associated with different applications to work together.
Extensibility is more complex and involves adding functionality to standard chatbots. “To do this you need to be able to make changes at a granular level,” Biran explains.
The SAP roadmap for conversational AI will see bots able to be developed from FAQ documents: this will be available by the end of this year.
Also in the fourth quarter of this year, we will see the first examples of multi-bots.
The next version of conversational AI will include smart disambiguation and misunderstanding support. “For instance, the same sentence or phrase can have different meanings in different systems.”
Among the methods that can be used to solve this are taking the application context into account, Biran points out.
“The first version will be naiive and ask the user what they meant. The second version will be smarter and able to resolve them automatically by using previous explicit actions.”
Emotion detection will become an important feature of chatbots, and an important part of boosting user experience.
“The starting point is experience: chatobots can boost experience and that is their value,” Biran says. “We want the experience to be boosted via the chatbot, so it’s interesting to know what emotions users have when interacting with them.”
Additional features that conversational AI will aim for in future include REST skills generation; the multi-bot as a hub to allow customers to connect non-SAP applications; bot delivery and extensibility; the enablement of marketplaces; and the ability to draw business insights from the data generated by chatbots.