Managing with machines

A chatbot is now available to answer ETH members’ questions about the IT services offered by the university. The IT Services team recently introduced the bot to help people help themselves. But IT Director Rui Brandao admits they still have some way to go. “In about half the cases, the answers are useful, but the rest of the time the chatbot misses the point of the question,” he says.

Infrastructure for research

The chatbot is one of the first AI projects in the field of ETH administration. “We use a number of rule-based systems, but they are not yet what I would call AI,” says Brandao. The chatbot, on the other hand, gets better over time by learning from user input and responses. Its primary purpose is to answer the questions on IT services that students typically have when they start their course. Chatbots are also used by customer service departments at Ikea and other companies, says Brandao, though he adds that most of these systems are at a fairly embryonic stage. IT Services has to provide reliable technology to thousands of people on a daily basis, he notes, so AI can only be deployed in isolated cases. “The systems we use on a daily basis need to be robust,” he explains.

One application in which Brandao and the IT Services team have opted for AI is a key piece of research infrastructure known as the Leonhard cluster, which is specially designed for big data analytics and machine learning. “It offers features unmatched by any other cluster in the world and is very popular in biomedical research and other areas,” says Brandao.

AI for business

And it’s not just researchers who are currently wowed by the potential of AI, says Stefan Feuerriegel, ETH Professor of Management Information Systems. “Companies will find that AI gives them a competitive edge in the long run. But it will be five years or more before that becomes visible,” he says, explaining that we are still in the early stages. “Companies are just starting to experiment with AI – and we’re here to help,” says Feuerriegel. He cites the example of AMAG, Switzerland’s biggest car dealer, which commissioned him and his team to define the most potentially exciting AI applications for the company and to launch some initial projects on that basis. Feuerriegel’s team is also helping online retailer Digitec Galaxus develop an intelligent system that will analyse customer behaviour on the website. The aim is to identify hesitant customers and provide them with additional information at the right moment in order to boost their resolve to go through with a purchase.

Feuerriegel argues that predictive analytics – in other words, data-based forecasting systems – are a promising area for AI applications, whether in marketing and sales, healthcare and insurance or logistics. AI can also help with traditional administrative tasks, as demonstrated by a recent ETH spin-off. The idea is that, in future, repetitive tasks such as entering invoices, checking delivery notes and processing expense receipts could all be carried out by machine learning algorithms. The key is to create algorithms that can read and process invoices and receipts even if they are not specifically available in a computer-readable format. The solution developed by BLP Digital is based on a combination of two technologies: image and text recognition. BLP expects to see interest from customers in all sectors where administrative processes consume significant resources. “We know that even processing a simple invoice takes an average of 8 to 12 minutes,” says Feuerriegel – meaning that AI could achieve major time 
savings in this context.

This sounds appealing – but what does it mean for jobs? Studies suggest that AI could lead to the loss of a good 20 percent of jobs in the administrative arena. That might seem like cause for concern at first glance, Feuerriegel admits, but he insists that AI will also provide plentiful opportunities for more interesting and higher-qualified jobs than we have today. “We’ll always need the human factor,” says Feuerriegel, noting that this trend will only unfold at a gradual pace, not a disruption that will change everything from one moment to the next. “We can’t purchase intelligent systems off the shelf as if they were smartphones,” he says. It is still a matter of developing the right solution for each individual product, and this process takes time.

Part of the digital strategy

ETH is steadily digitalising processes for resource and business management, human resources management and services for students. “AI is part of our digital strategy,” says Robert Perich, Vice President for Finance and Controlling at ETH. That puts AI for administrative processes firmly on the ETH agenda in the context of continuous organisational development and digitalisation, for example within the framework of the rETHink project. His colleague Paul Cross explains what this means: “Our aim is to take a 360-degree approach to digitalisation and ensure that we have the kind of solid foundations for AI that allow us to properly align people, processes, data, systems and governance.” The idea is to work closely with ETH researchers and harness existing relationships with experts in machine learning, natural language processing and other areas of AI. “We can draw on world-class expertise at ETH,” says Cross. In return, he argues that the process of digitalisation can supply researchers with use cases that they can put into practice. Cross is confident that AI will become a valuable tool for administration at ETH within just a few years – a tool that will offer plentiful advantages to students, staff and other broad groups of stakeholders.