Pension provider, Velliv, has entered into a collaboration with Danish artificial intelligence (AI) company, 2021.AI.
The partnership follows a greater focus from the European Union (EU) after it published its Data guidelines and Ethical Guidelines for Trustworthy AI model development in February 2020. The purpose is to ensure consumer rights as well as responsible processing of consumer data now and in the future.
Velliv will use 2021.AI’s Grace platform, a leader within AI Governance. Grace standardizes processes and workflows across areas such as data management and AI model development to secure complete transparency, traceability, and audit trails.
“With the Grace platform, we have a head-start supporting the EU’s guidelines to improve citizens’ trust in organizations’ use of data and AI. We can ensure full traceability and documentation to support the new EU’s guidelines for those working with data and AI, what they are working on, and not least, who is responsible for the work. Grace supports this via a combination of complete audit trail, history, and Impact Assessments, which is a range of specific questions and considerations, that organizations must go through in the AI model development,” says Mikael Munck, CEO, and founder of 2021.AI.
The pension provider is ready to further develop AI models after using robots to enhance customer service for several years. They said AI is a natural extension to this.
“Our customers’ confidence in us is our most important asset. Until now, we have been reluctant to develop and acquire AI due to security and ethical considerations,” Velliv CDO, Christine Hunderup, said.
“We have sensitive data that we need to secure for our customers. The EU’s data, AI guidelines, and 2021.AI solve that challenge, which now gives us the right basis to accelerate the development of AI to ensure a better and more targeted customer service without compromising on security and ethics.”
Christine Hunderup highlights the Grace platform’s completeness where AI model development, data, and AI governance, guidelines, and traceability are consolidated into an integrated solution.
Implementation of AI and data governance has been a deficiency, which many working with AI have tended to overlook. Such as the case of Apple’s credit scoring model, which automatically assigned far higher credit ratings to men rather than women.
“So far, AI has been characterized by innovative experiments rather than a focus on practical implementation, considering data ethics and data security. However, we now have the opportunity to document both our general governance and ethical considerations directly in our models, as well as the continuous tracking of all activities done carefully and automatically. This happens while data remains in Velliv’s systems, which has been crucial for us to accelerate the implementation of AI now. It is now safe for our customers and us,” says Christine Hunderup.
The work in the EU has also aroused interest in The Confederation of Danish Industry (DI), where Director, Lars Frelle-Petersen, welcomes the initiatives and states: “The EU guidelines in this area are a good start, as companies now have something more concrete to relate to. Our assessment is that the EU guidelines go hand in hand with the implementation of a data and AI platform. Partly to validate that guidelines are followed and in the same context, that the highest level of data security is implemented, which the EU has been fully aware of. This is a step in the right direction for citizens and customers. Danish companies will advance in the field of AI, which supports our already strong position in the market for robots.”
Christine Hunderup fully agrees with Lars Frelle-Petersen. Implementing a platform that administers compliance with AI governance and data security will strengthen Danish companies’ position:
“Now we have the right platform, and unlike some of the pioneers in the field, we will not have to stop or start over, when implementing the new regulations and rules. We expect to have the first 2-3 models in production during the first half of the year. For AI implementation, it made sense to slow down and wait for a complete solution for data and AI accountability.”