Christian Guttmann, a German-Australian scientist and entrepreneur, has more than 25 years’ experience working for the world’s largest U.S. and European IT companies (including IBM, HP, Ericsson, and British Telecom). He also founded and worked at several AI-first startups. He is currently Vice President, Global Head of Artificial Intelligence and Chief Artificial Intelligence and Data Scientist at Nordic software and IT company Tieto, and Executive Director of the Nordic AI Institute, (NAII), a global alliance dedicated to using AI and machine learning for social and economic prosperity. Guttmann is also an adjunct professor at the University of New South Wales in Australia and a senior researcher at the Karolinska Institute in Stockholm, Sweden. He has written and edited seven books, released more than 50 publications and holds several patents. He recently spoke to The Innovator about how companies should think about implementing artificial intelligence (AI).
Q : What should companies be thinking about when they implement AI ?
CG : There are three factors that drive the successful adoption of AI : one is the strategy, which means a high level of commitment from the leadership and a clear understanding of how the company wants to use AI to reposition itself in the emerging AI era, such as where it wants to compete and what capabilities it will need. Companies have to be very rigorous about this. The CEO needs to be ready to present the AI strategy to the board and articulate ‘this is how we are proceeding and this is what we will be known for in 2022.’
The second factor to consider when you are building your AI and your products and services is how this will change the company’s business processes. You can’t just build an AI product or service and drop it on the employees or customers. You have to think about how the business will run differently and how a company needs to change its routines. The other part of the business impact is trust in AI: you need your employees and customers to trust that AI will do what it is expected to do.
Finally, you need to make sure your company is data and IT ready. You need to determine how quickly you can access the relevant data or computational power to run the planned AI services. This is not trivial. Data native companies have the purpose of good data written in their genes but if you are a forestry company or a container ship company that has never had data in their DNA it is likely that you are not sufficiently prepared. The average company has dozens if not hundreds of databases that are not connected. For a proof of concept (POC) ad-hoc data collection might just do the trick, but if you want to operationalize the AI services sustainably, you need the IT infrastructure to work smoothly so you can successfully use AI to optimize and predict. For that you often need to have a data hub and this needs to be created before you make the AI processes part of the core process of your organization. If this requires a major change then you need to bring this to leadership from the beginning.
Q : What are some of the other major challenges with AI ?
Trust is a big issue not just with employees but also with customers. Appoint a chief of customer experience. It is paramount that customers are comfortable and have a very positive experience. Then the question is how do you create that link with what you are producing? Think about what policies you will put in place.
Another issue is how do you make a business case for AI? The money earned from AI services in the entreprise world is still not very high. Internet and ecommerce companies are having the most success with AI but in the enterprise world things are going slower.
And, finally, it is still difficult to predict the extent of disruption. We are at a point in time that is similar to the development of the Internet in 1996 so you need to keep an open eye.
Q : What sort of safeguards do companies need to put in place to guard against bias in AI and damage to their brands ?
CG : It may be necessary to do a risk factor analysis. There is a risk that there will be unforeseeable consequences and this requires setting the expectations for the stakeholders. While the benefits are very clear, AI is not a fail proof technology and setting wrong expectations could severely damage your brand. In some cases, there probably needs to be some sort of insurance system in place. Even if a company follows certain guidelines in training an AI, there is a risk that the technology will make mistakes just like humans do. In some cases, increasing trust in AI may mean to be transparent about the way the AI was built, what data was used and the tests that were run. You will need a unit to fix and analyze problems, an AI ops that deals with downstreaming AI products into the customer environment. You may need continuously automated AI quality checks and to additionally test for robustness, predictability and accuracy.
Q : Are there industry standards that companies can adhere to ?
CG : There are groups working on this including the Institute of Electrical and Electronics Institutes (IEEE), International Standards Organization (ISO), Information Industry Technology Council (ITIC) and the Software and Information Industry Association (SIIS) as well as the U.N Centre for Artificial Intelligence and Robotics, the OECD and the European AI High Level Expert Group and the EU AI Alliance. But don’t sit on your hands waiting for the ecosystems or governments to create these guidelines. It is in your own self-interest to care about your brand and trust to put your own policies in place. This is all evolving. We need to create a well-functioning framework but in the meantime each company has to work on its own policies and work with its customers and stakeholders to achieve the highest possible level of trust and customer experience.
Q : How is your company, Tieto, dealing with these issues ?
CG : AI is a multi purpose technology that will be used in every business. We assessed that there are some 300,000 use cases and business opportunities. We had to decide which of those we would target. Finance AI is one, business service AI is another and we also focus on AI manufacturing. For example, in the public sector, business services can be chatbots that improve citizenship services. We are known to be the first choice for customers in intelligent service delivery for the Nordics. Chatbot technology allows you to create a relationship with individual customers, to maintain that relationship and establish trust. The message we want to convey is that our tools are trustworthy and conform to the highest possible ethical standards. This is a strong value proposition in the Nordics. Our pitch is that we are using known methods to make the development and use of AI as fair as possible.