Three Female Executives Discuss Generative Artificial Intelligence: ‘A Significant Call to Action’

To delineate the concept: artificial intelligence, or AI, is not a new phenomenon. What is fundamentally different about GenAI?
Qiu: ‘Classic AI is discriminative rather than generative. It focuses on creating algorithms based on existing data to classify or predict output. GenAI ‘trains’ data to generate new human-like content, such as translating text into new text, or text into an image, or an image into a video, for example. The power with which this happens is enormous. That is what makes it revolutionary. It is truly a milestone in the development of AI.’
Moerel: ‘Classic AI is trained with domain-specific data to perform specific tasks within that domain, such as a recruitment tool. GenAI is trained with unlabelled data from all possible sources and can handle all kinds of tasks simultaneously. Another significant difference is user-friendliness. As a consumer, you input a prompt, like a question or command, and you get the output in no time. Many people use it more or less like a search engine that not only looks for relevant information but also generates content. I see it as a real game-changer.’
Kuijt: ‘Moreover, its scope is much larger due to the way it is being marketed. For example, the language model ChatGPT became widely available to consumers while it was still in development. We have not seen that before.’

Let us make it concrete. How do you apply GenAI in your own work or organizations?
Kuijt: ‘IBM has developed the Watsonx platform and its own foundational models to integrate GenAI into business operations. We develop basic models for organizations that want control over the data feeding the model. For example, we collaborated with the space agency NASA to develop HLS Geospatial FM. It is a model for earth observation data exclusively fed and trained with data collected by NASA using satellites. After the launch, it was made available as open source. Its value for science is enormous. You can track changes in land use, monitor natural disasters, or predict crop yields, for example. Initially, it takes a considerable amount of work to gather the right datasets, but once the model is in place, the applications are not that costly. That is the beauty of it.’
Qiu: ‘Signify already uses GenAI to speed up software coding; in some cases, we reduce the required time from weeks to hours. In our customer contact centers, we use GenAI to accelerate responses to customers. Where an employee needed an average of 2.5 minutes, the model takes around 8 seconds. It still needs to be checked by a human, which takes about 20 seconds, but it is a significant acceleration. Moreover, since the model is trained with our own data and fed with all new solutions, the answers become more accurate over time.’
Moerel: ‘In my advisory work, I mainly see large-scale applications in call centers. In fact, I do not know any major company that is not involved in this, leading to many people worrying about their jobs. This concern is valid, but for the wrong reasons. They do not need to fear being replaced by GenAI; they should be afraid of being replaced by someone who knows how to handle GenAI. And that applies not only to call center employees.’

What does this mean for the future of work?
Qiu: ‘GenAI is already leading to staff reductions in certain parts of the business. Unfortunately, we find that many people are afraid to work with it because they do not understand it well. This is particularly challenging because GenAI continues to develop rapidly. The learning curve is enormous, and the question is how to implement knowledge about it throughout the entire company.’
Kuijt: ‘I see the future of work changing for both people on the shop floor and for managers and executives. Decisions are now much more data-driven and therefore more transparent. GenAI will elevate the use of data to an even higher level. It truly is a call to action. Start thinking now about how GenAI can help lighten or improve your work.’
Moerel: ‘This applies to virtually all functions. Take the corporate secretary, for instance. With GenAI, efficiency gains can be made in terms of reporting meetings to the board, supervisory board, and the general meeting of shareholders. You can also compare all annual reports of companies in your sector and generate suggestions based on that, such as identifying certain risks or developments. The possibilities are not yet fully exploited. When I have discussions about further education in this area, it is always about someone else, never about the people present in the room.’

Does that apply to executives as well? Are they adequately equipped for this topic?
Kuijt: ‘I clearly see attention from executives regarding its development and its value for their organization. Technology used to be something for a group of tech enthusiasts in the IT department. Now, executives can apply it themselves; their children use it… that enthusiasm drives the discussion beyond compliance aspects, focusing also on opportunities and possibilities. Naturally, the upcoming European AI legislation also demands attention from executives.’
Moerel: ‘However, leadership is not sufficiently on top of it and there is insufficient fundamental thinking about the implications of GenAI for the mission and strategy. It permeates your entire business operations and all departments, so every executive should educate themselves in this area. Otherwise, you will not even have a common language to discuss the new possibilities and risks, and you will not know which direction to take when unexpected issues arise. I see Dutch executives asking their legal department for a report on all the risks, and there are quite a few. Dutch companies are relatively more hesitant to use GenAI because of all the risks. They prefer to wait until the European AI Act clarifies where they stand. There is almost a sense of relief: luckily, we do not have to innovate yet. I am generalizing, of course, but there is some truth in it. Yet I think that nobody can afford to wait until it is completely safe and regulated. Otherwise, others will surpass you. Experiment and explore how it can be used, without becoming dependent on it and closing the door to reverse what has been done. You need to learn to walk before you can run. That is what it boils down to.’
Kuijt: ‘At the same time, you should not deviate from your normal methodological way of working just because GenAI is new and exciting, and you do not want to fall behind. Start with questions like: what do I want to build? What purpose do I want to use it for? Is this really the best technology for this purpose? Where do I get my data from? Have I built in checks to address bias or other risks? Are the users trained well enough to work with it? Do I have enough subject matter experts in-house? Then you have to monitor it and make sure there is a way back if the model does not do what you aimed to achieve.’

Certainly, there are plenty of risks associated with using GenAI, such as in the areas of privacy, copyright, accuracy, ethics, bias… In Europe, legislation on AI has been in development for over two years. What do you expect from it?
Moerel: ‘The Commission’s proposal for the AI Act predates the introduction of GenAI, so a new category had to be created in the law specifically for these foundation models. Negotiations on the AI Act are now in the final stages, and it is expected to be adopted before the end of the year. Meanwhile, the Dutch government still needs to determine its position on AI. In that sense, as a country, we are lagging far behind. The AI Act certainly has its criticisms – just like the GDPR – but it provides good basic principles to assess algorithms for accuracy and bias. Governments and companies will have to be accountable for this.’
Qiu: ‘Elsewhere in the world, people are eagerly awaiting the European AI Act. They know that legislation in Europe often takes a long time to materialize, but what materializes is usually good and directive. Signify is a global player, but in certain cases, we take European regulations as the standard because they are the most comprehensive and often the strictest. As long as the AI Act is not in place, we remain cautious. For us, it would be very helpful to know what is allowed and what is not, but there is a lack of unity and decisiveness at the European level. The problem is that Europe itself does not have major tech players like Amazon, Google, Microsoft, Tencent, or Alibaba in the world of the internet of things, so it lacks the influence to lead in technological trends. This is now also happening in the field of GenAI. In the area of Large Language Models (LLM), we have Bard from Google, ChatGPT from Microsoft, Ernie Bot from the Chinese Baidu. Such tech companies are at the forefront of innovation and initially set the rules. There should be more political support to further develop the European initiatives that do exist.’
Moerel: ‘We may see ourselves as the world’s regulator and be proud of the ‘Brussels effect’, but bottom line is that referees never win the game. You have to play and innovate yourself; otherwise, you have a competitive disadvantage. Cloud providers now deliver vital infrastructure, but almost all European data is still in foreign clouds, making it less available for European innovation with AI. This dependency is not good for Europe.’

We might not be able to fully anticipate developments around GenAI, but still: where might we be in five years? What will it have brought us?
Kuijt: ‘I think there will be specific foundational models tailored to specific areas in all sectors, and people will really use them, also because people will demand it. At some point, it will become normal for, say, doctors and lawyers not only to rely on their own expertise but to use all possible sources. I also hope that the technology becomes truly inclusive, so people want to embrace it.’
Moerel: ‘The AI Act applies to all types of AI. I expect there will be specific standards, rules, and certifications for specific applications. Just like we now have endless checks specifically to ensure that cars or airplanes are safe. For the application of AI, it is still like driving in a car without brakes. More and more rules will be established to get this under control. It is unacceptable to use a product where you do not know if the output is safe or correct.’
Qiu: ‘Many technologies come and go, but GenAI will not disappear. It will have its ups and downs; the impact is significant and enduring. That is exciting and a bit uncertain. But is that not the beauty of life: learning to deal with the unexpected?’

Interview by Elisabeth Philipse, Partner at Kearney, and Ingrid Reichmann, Co-Owner of Partners at Work. Published in Management Scope 10 2023.