Sandro Saitta, PhD: “Every AI Project is an R&D Project”
This interview has been conducted in partnership with Business School Lausanne and is a part of the GTF Technologies for Governments Lab report.
Sandro Saitta is the founder and CEO of Viadata and the former Head of the Industry Unit of the Swiss Data Science Center (SDSC). He worked with companies such as Richemont, Logitech, Merck, Adecco and Firmenich to help them adopt innovative data solutions. Sandro is a lecturer at Business School Lausanne (BSL) and in the Certificate of Advanced Studies (CAS) in Data Science & Management at HEC Lausanne. He is also a co-founder of the Swiss Association for Analytics. He holds a PhD in Computer Science from the Ecole Polytechnique Fédérale de Lausanne (EPFL).
“If an organization consistently finds itself in the position of persuading stakeholders of AI’s value, it could indicate a less mature state of AI adoption.”
GTF: How do you define AI adoption, and at which point can you say that an organization has successfully adopted artificial intelligence?
Sandro Saitta: AI adoption is a multifaceted concept. The approach that I like most involves categorizing AI adoption into various levels, similar to a scale with five distinct rungs — basically, from nothing at all to an AI-driven organization.
To gauge AI adoption, we employ assessment tools and methodologies that provide organizations with a comprehensive understanding of their position on this adoption spectrum. An important point for me personally is that AI adoption is a dynamic journey, with opportunities for improvement at every stage. Most organizations don’t need not aim for full AI-driven status. Progress can be made incrementally.
A good informal marker of AI adoption that I use when I work with organizations is the necessity to continually justify AI projects to senior management. If an organization consistently finds itself in the position of persuading stakeholders of AI's value, it could indicate a less mature state of AI adoption. In contrast, data-driven companies have successfully integrated AI into their culture, eliminating the need for perpetual justification. However, traditional companies often encounter resistance and require significant efforts to sell and implement AI projects.
GTF: Do you think public sector organizations, such as governments or municipalities, are culturally closer to these traditional companies that require internal explanation and selling processes?
Sandro Saitta: Drawing from my experience, public sector organizations often exhibit cultural similarities with traditional companies when it comes to AI adoption. For example, in both you often see a low turnover, some people are there for several years and they like working in a certain way. And when you bring AI, you bring change, and they might not be very open to new technologies, trying new things and new ways of thinking.
Implementing change, especially when introducing cutting-edge technologies like AI, can prove challenging. Moreover, public organizations handle sensitive data, such as citizen or patient information, further complicating the adoption process — for many new ideas to work you need new laws, new regulations put in place, etc.
“Individuals who exhibit curiosity about AI and are willing to explore novel approaches tend to be more successful in the adoption of AI initiatives”
GTF: Do you think this is also a cultural difference, in addition to the different managerial structures? Is the cultural aspect a significant challenge in public sector AI adoption, and what should an ideal culture for AI adoption look like?
Sandro Saitta: Indeed — because before anything else, one must be able to accept change. And the leadership must be capable of leading that change. And this is a cultural problem. Thus, cultural considerations play a pivotal role in AI adoption, regardless of whether it pertains to the public or private sector. Organizations must be not only willing to embrace change but also take a leadership role in driving it. Embracing change involves reskilling employees, adapting existing processes, and wholeheartedly adopting new technologies. It's equally important to foster a culture that accepts the inevitability of failure, because any AI project is, essentially, an R&D project. People and organizations must understand that. You will never know in advance if it is going to work the way you want it to, how much value will it bring, and so forth.
Here enters another essential cultural element: a curious mindset. Individuals who exhibit curiosity about AI and are willing to explore novel approaches tend to be more successful in the adoption of AI initiatives. When dealing with senior employees in public sectors, offering coaching and tailored training can be invaluable in bridging the gap in understanding and implementing AI technologies. I think, leaders in the public sector are the ones that need such personalized coaching the most.
“if you want to bring change in a private or public organization, you have no choice but to work on an individual level. ”
GTF: To address this problem, can you educate long-standing employees in traditional sectors about AI, or should organizations wait for generational turnover?
Sandro Saitta: The education of long-standing employees is not only feasible but often, probably always, necessary. To accomplish this, organizations must highlight the individual benefits of AI adoption. This is really important: if you want to bring change in a private of public organization, you just have to work on individual level. Demonstrating how AI can enhance daily tasks and emphasizing the "what's in it for them" aspect is crucial. What can this or that particular employee gain from all that? If this is not addressed, the adoption rate will be very low.
So you need to run hands-on sessions, show and implement no-code solutions, and personalized coaching as I have suggested, can render AI concepts much more accessible. You need to be able to counter those people who have this “I have been doing it for years and it worked just fine” attitude. They need to understand that it will not work “just fine” anymore because of the scope of the change AI brings. No-code solutions can be eye-openers here, with people creating their simple models and really learning how it works. The objective is not to transform them into data scientists but to provide a foundational understanding of AI and how it can facilitate their work.
GTF: Based on your experience and your thinking, can you share two or three general rules that organizations should follow to successfully adopt AI?
Sandro Saitta: I can think of a few things for sure. I would say the main universal principles revolve around what we call data literacy, understanding of how to read, write and communicate data. Here is how I would explain the three main ones:
First, promote data literacy. Organizations should invest in educating employees at all levels about data. This encompasses the ability to read and communicate data effectively, understand basic statistical concepts, and interpret data visualizations. A solid foundation in data literacy is a prerequisite for any AI initiative. I cannot stress this enough: everybody in the organization should be able to read charts, build dashboards, avoid typical traps, and so on.
Second, encourage understanding of AI. Ensure that all personnel possess a fundamental grasp of how AI functions and its inherent limitations. This understanding is crucial for managing expectations and building trust in AI systems. Here again, it is important that everyone, even on the most basic level, has some understanding of statistics and data visualization. Once again, it is not about transforming all workforce into data scientists, but explaining both the potential and the limits of AI to everyone.
Third, emphasize data quality. It comes again to the question of understanding how AI works — that without good data, it is useless. Employees should be aware that AI heavily relies on high-quality data. I have seen so many cases in different places where projects don’t work despite all the data scientists, all the tech, all the processes being in place… just because the person who is doing the work, feeding the data, is not feeding the right things, leaving a default value somewhere, or leaving it blank. Most likely, this person is not feeding the right data because they just don’t know why they should do that. Demonstrating the value of accurate data and motivating employees to contribute to data quality are essential components of a successful AI adoption strategy.
So you see, adoption really comes down to culture and education, because AI is such a broad phenomenon — be for a corporation, or a government.