On excellence, the perfect timing for AI innovation, data economy, and why being flexible matters

We're living in what could be the most important century. And at the center of this era is Artificial Intelligence and data science, reshaping all aspects of our lives.

And as we're all adapting to these changes, as individuals and companies, we're turning to those who steer this revolution from the front lines. And among them we find Visium, an AI and data company that has been guiding organizations through their AI journey since 2018.

In this exclusive interview with Visium's founders, Alen Arslanadic, Timon Zimmerman, and Matteo Togninalli, we explore a range of topics, from the challenges and opportunities of AI adoption in various sectors to strategic insights into building a customer-centric, authentic AI and data company.

As they navigate the complexities of AI innovation and have played a key role in applying these technologies across diverse industries, their insights serve as a guide for those looking to understand what increases the chances of successful AI adoption and how to approach it altogether.

We can't start this conversation without covering one of the most important topics in Visium, and that's the company's values.

Excellence is a value that has been embraced in Visium since the very beginning. I won't ask what it means to you, I'd rather like you to share an example where "excellence" was key in solving a client challenge.

Timon:

We've noticed that often, data science teams struggle to showcase the value they bring to the company. It's a common challenge that we found to be true in many organizations across different industries. They find it difficult to present the success of their data initiatives to the company's board and top management. And that leads to the executive team often not being aware of what the data science team is up to. 

And I believe we're showing excellence in our client relationship management as we've taken these opportunities to help them shine in front of their leadership. We're helping them excel by bridging the communication gap, step by step, with every opportunity to share the impact of data science.

Visium also has a deeply rooted culture of continuous learning. How does it translate into benefits for customers?

Matteo:

The main advantage for our clients is that our team's lifelong learning value is encouraged in all roles, promoting ongoing knowledge sharing. So when our clients work with a Visium engineer, they access not only that individual's expertise but also the collective knowledge of our entire team.

Timon: 

I agree, and I think we also help our clients navigate the overwhelming information in AI, tech, and data. It's a lot to handle, even for those with a technical background. So, I can't imagine how difficult it must be for someone non-technical to distinguish between big news and real technological advancements from mere marketing buzz. The term "infobesity" captures this challenge well. We help our clients stay focused on what really matters amidst all the fuss and noise.

Alen:

One of the biggest advantages I see is that our team has consistently delivered innovations to our clients years ahead of their time turning them into competitive advantages. For example, already back in 2018, we developed a Generative AI solution for chemical formulation for dsm-firmenich. Then, in early 2020, we applied transformer architecture to a computer vision task in digital pathology for Roche. It was only two months before that particular approach appeared in the scientific literature and three years before it revolutionized the world through Large Language Models (LLMs) and ChatGPT.

Are there any principles or practices you're refusing to compromise on at Visium?

Alen:

Being ethical and responsible with our AI practices, without a doubt. At our core, we care about pushing forward ethical and responsible AI, and we always select use cases with that in mind. 

Timon:

Yes, we have had at heart this ethical and responsibility aspect since the beginning. And although it has sometimes been challenging to draw the line, we have maintained our commitment to ethical and responsible AI. And we created different mechanisms to support that, such as the Ethical Committee. It may not be flawless, but I don't think perfection can be achieved for such a topic.

Matteo:

Mm, I like the way it was originally phrased: "To do the right thing even when no one is watching". It's more broadly linked to integrity rather than ethics and responsibility, but it's a sound way to think about it.

What qualities do you look for in employees, customers, and partners? And is there an overlap in these desired traits?

Alen:

We seek team members who strive for excellence, embrace responsibility, thrive on constructive feedback, have a relentless drive for improvement, possess a strong sense of justice, and can show deep empathy for colleagues, clients, and the world. For partners and customers, we look for trailblazers eager to shape a great future and lead in their industry. And in the end, shaping a great future relies on the things I enumerated above, right?

Matteo:

I would definitely say ambition and moonshot thinking.

Timon:

I would say curiosity. We want clients to be curious about AI, data, data engineering, cloud computing, and technology in general, and understand how these can help them. And the same applies to our employees. We hire people who are curious and passionate about what they do, regardless of their role within the company.

Visium is working with companies across 5+ industries, what do they have in common concerning adopting AI?

Alen:

We categorize companies by their AI maturity, offering tailored solutions to help them advance to the next stage. Generally, the challenges of building data science teams and running successful initiatives are similar across industries, which is why we work across various sectors. While industry specialization and domain expertise can accelerate AI success, they're not always influential factors.

Matteo:

We mentioned curiosity before, and I think it's something they have in common. But there's also this focus on ensuring that they're not undertaking toy projects but rather initiatives that can truly transform their business at the core.

Timon:

And they have a strong "why" on AI and data initiatives, which is a strong differentiating factor.

Where does the strong "why" come from? How come some companies have it while others are still struggling with it?

Alen:

Innovation, particularly in AI, doesn't yield immediate results and is often misunderstood. Until recently, companies didn't see AI as a critical capability. So, without a pioneering spirit or a digital leader championing it in-house, companies struggled to find a compelling reason to adopt AI. But also, adopting AI is challenging, and many initiatives fail to reach production, leading to more skepticism and disillusionment. As AI became more mainstream and companies recognized its potential, skepticism started to diminish, accelerating adoption. With less internal resistance, more companies are now reaching a tipping point where they have a strong rationale to advance AI.

Timon:

And it mostly boils down to executive and board-level sponsorship, as the strategic "why" for AI initiatives typically originates from the top. While employees can certainly inform and encourage leadership, the ultimate decision-making and strategic vision for AI investment must come from the executive team.

Matteo:

Yes, and it's quite interesting to see how it's always present, but always different, right? And it's always related to the company's core strategy and mission. So if we think about dsm-firmenich - it was all about empowering the creativity of the perfumers, if we think about one of our CPG clients - it was all about improving the customer care for their machines. So the "why" is not necessarily technically driven. On the contrary, it's very much linked to the mission and the vision of the company.

In your view, what constitutes being "too early" or "too late" in adopting AI technologies?

Alen:

Being "too early" means investing heavily in an immature technology, resulting in underperforming software. On the flip side, being "too late" is when a company hasn't even initiated a small-scale pilot of a core AI-enabled capability, while many of their competitors have already successfully implemented it.

Timon:

I think being "too early" to adopt AI might be less about timing and more about expectations.

If you don't have the right data, cloud, and IT infrastructure foundations - it doesn't mean you cannot do AI, but you need to have the right expectations. At first, it might be more about creating a flagship proof of concept to justify investments in a data platform and then setting out to do AI. 

Regarding being "too late", I don't think it's too late for any company yet, but it will soon be. If you haven't run any AI initiatives yet, and you haven't started working on your AI and data strategy – I would be a bit worried, yes.

Matteo:

I think being "too early" isn't exactly a disadvantage, but it might mean that it takes a bit longer to figure out the most effective way to operate AI and data teams. But you'd still be early compared to companies that follow and learn from your mistakes. And it's potentially a larger initial investment. But it still allows you to be ahead of the pack, if you're consistent with it.

How does Visium navigate these timings for clients?

Timon: 

What we mostly help our clients with is realizing that there is no perfect moment to start working on their data maturity, data platform, and run AI proof of concepts. They have to take some risks. And it really takes a good chunk of willpower from top management to push this type of initiatives, because if they wait for the stars to perfectly align, it might never happen.

Alen:

We usually begin by categorizing potential use cases according to our clients' technological maturity and expected ROI, among other criteria, to prioritize the right initiatives and help them get started.

Matteo: 

And we have blueprints for nearly all types of configurations, right? If a client is eager to implement AI but lacks the foundational infrastructure, we can help in establishing the right setup from a data or cloud perspective. If, on the other hand, they have a robust data architecture but lack an AI vision, we can help design a roadmap around multiple AI use cases.

We offer a broad palette of services that can be perfectly tailored to the state of any company. Whether it's too early or too late in comparison to the rest of their industry or their internal readiness, we remain flexible and committed to serving them in the best way possible.

Reflecting on this quote from Tao Te Ching: which is mainly about adaptability, flexibility, and resilience, how does this philosophy apply to companies adopting AI? And can you provide an example of how being adaptable has benefited a client?

Timon:

What's interesting about this quote for companies and AI in general, and I'm thinking a bit out loud here, is that flexibility actually helps build resilience. It might be more about adaptability than flexibility, but both seem to work hand in hand in building resilience. Take COVID, for instance: the companies that pivoted the best and adapted their working methods were those that showed the most resilience during that chaotic phase.

It reminds me of another quote, I don't remember which book it was: "Companies are like sharks, they die if they stop swimming". And I think it's very true. If companies stop adapting and constantly pivoting, that's how they die and become brittle.

Alen:

It makes me think of one of our projects. We built a tool to structure data from a public annual report for a financial services firm. During this development, significant advances in the field enhanced the state-of-the-art for our task at hand. This led us to replace the solution's core intelligence with newer algorithms, resulting in notable performance improvements. Developing AI solutions amid rapid field advancements requires constantly being ready to adopt newer approaches for the same challenges, unlocking increasingly powerful solutions.

Matteo:

Another example comes to mind. We developed an algorithm for one of our client's projects, which was working well and ready for deployment to their internal users. However, what was missing was the integration of the algorithm into the rest of the workflow. So, instead of 'striking the sails,' what we did was: our team continued to refine the model while the client simultaneously worked on these more fundamental aspects.

So it's a matter of staying committed to the original objectives, but not being too stubborn and remaining a bit agile on the way as difficulties arise.

How can you make sure that you don't have a false perception of resilience? How do you keep it in check? Even from a technological point of view.

Alen:

You attain resilience by persisting until reaching a successful outcome, despite facing many challenges and obstacles. As long as you continuously make an effort in the face of adversity, resilience exists.

Timon: 

And there are so many examples, especially during COVID, where complex IT systems made it hard for companies to adjust quickly. These are the same systems that created this perception of safety and resilience every day precisely because of their complexity.

But adding additional complexity doesn't make the solution better, more resilient, more robust, and more scalable, but rather the opposite. Sophistication is rarely a good thing when you're coming up with a new process or technological solution.

Matteo:

Mm, yes, that reminds me of the quote, "Simplicity is the ultimate sophistication", which captures the essence quite nicely.

With AI rapidly evolving, how do you keep up with all the new developments?

Alen:

I am fortunate enough to be in a position where I interact daily with global data science leaders and our incredibly passionate and brilliant engineering teams. This exposure keeps me abreast of the latest and greatest in AI. Coupled with reading in my free time and subscribing to a few insightful newsletters, I find myself almost effortlessly staying on top of the key advances in the field.

Matteo:

There's probably never been a better time to learn things. There's a wealth of information available and just as many ways to consume and acquire knowledge. So it's not the material or the tools that are missing. The key point is choosing where to focus one's energy. I try to find a common ground between my fundamental interest in AI and what can be beneficial for Visium and our clients. And the rest it's just time management skills. 

Timon:

I've learned to accept that I won't be able to keep up with the depth of knowledge that our engineers have in terms of implementation, technical details, and the math behind it, nor be as technically acute as I used to be as an individual contributor and engineer. So I just stay genuinely curious about what our engineers are learning and doing. So when there are new tech developments I usually talk about it with 2-3 engineers from our team to get a grasp of what they thought and understood. I think that's a good way to stay abreast of new developments. 

What are a few emerging technologies that you're excited about in AI and data?

Alen:

I find decoding the human brain to be one of the most impressive emerging applications of AI. Today's AI can already translate brain MRI scans into thoughts and images. While it may sound scary and intimidating, the potential to augment humans into super-intelligent beings is just around the corner, about a decade away. This milestone will redefine what it means to be human and further accelerate the exponential pace of progress. And I'm very excited to see how it evolves.

Matteo:

I think what's exciting, yet still somewhat unclear, is what will happen in the application layer amid all the LLM craze. So basically, the different ways in which these models can be tailored to specific use cases. There are tons of companies launching in this space. But the most fascinating part is that we don't know what will emerge to drive the next major technological wave. It's quite thrilling to be surprised in that regard.

Timon:

What I'm really curious to see unfold is the data economy - how companies engage in buying and selling their data in data markets. There are the obvious ones, like Bloomberg, but it will be interesting to watch how it unfolds for other companies too, even in industries such as Healthcare.

If Healthcare companies start developing business models, will they sell their models to other Pharma companies, or will they continue to guard them as they have historically, given their immense value and confidentiality?

I think more companies willing to sell their data to competitors will emerge. Which implies a significant mindset shift. Today, the idea of a CPG company selling manufacturing data, like machine specifics, seems far-fetched. But soon, this could be common, with companies realizing the mutual benefits of sharing data with competitors, creating a reciprocal exchange of information.