Home PublicationsData Innovators 5 Q’s for Tuomas Syrjänen, Chief AI Officer of Futurice

5 Q’s for Tuomas Syrjänen, Chief AI Officer of Futurice

by Eline Chivot

The Center for Data Innovation spoke with Tuomas Syrjänen, co-founder and chief AI officer of Futurice, a company based in Finland that offers analytical tools to various sectors to help them in their digital transformation. Syrjänen discussed how digital transformation is about using AI and data as much as creating cultural change and increasing connectivity between people and the business.

Eline Chivot: Futurice has transitioned from being a tech consultancy to becoming an AI and data-driven business. What was the strategic thinking behind this evolution?

Tuomas Syrjänen: When we started Futurice 20 years ago, we first found success by having the right technical skills and offering a modern way of doing things fast and properly. That carried us roughly through the first 10 years. Around 2007 we realized we needed to bring more value to clients and figured that this required building additional skills such as design, business, and data skills. We’ve been selling data and AI for the last 10 years, but nobody was buying that—at least not for the first six years. We grew fast and realized we needed to bring company culture into the core of Futurice. By this time, culture had become important to our clients as well. They started hiring us not only because they wanted the technical, design, and business skills, but they also wanted to adopt parts of our culture. One of our core success factors is based on this combination of having different in-house skills and harnessing culture—both internally and externally—and bringing that to our clients.

Now we see a need for a similar step to ensure we are still relevant in five years’ time. This means leveling up how we create value for our clients, improving access to our organizational knowledge, becoming more client-centric, and further strengthening our culture—with data and AI.

We’re not replacing anything from the past. We’re incorporating these new ingredients into our success formula. This is why we now talk about a transformation via data and AI, and have become a data and AI-advised company—or, as we call it, a “connected company.” As companies grow, they can easily become siloed and disconnected on many fronts, but with data and AI we can reconnect its parts, processes, and people. We can do this to our own company, and we can support our clients on the same journey.

Let’s take a closer look at how a “connected company” compares to others. In the classic, traditional organizational paradigm, managing uncertainty is done through a plan, decision-making is hierarchical and top-down, the team is large and functional, metrics include revenues or return on investments, and value is created through products. These organizations need to coordinate large numbers of people. The growth model that has emerged over the last five to 10 years is the innovation organization, which aims to harness people’s brainpower. Uncertainty is managed via making experiments, decision-making is bottom-up and characterized by more autonomy, teams are smaller and more cross-functional, and digital services are the source of value creation. In the paradigm we’re moving towards, which involves connected companies, data and AI are key. Managing uncertainty is about simulations, decision-making is data-driven, the team is modeled and optimized, metrics include the realization of probabilities, and data business models create value.

When we started our transformation program, we thought data and AI were about automation but over the years we’ve learned that data and AI in an organization are about connectivity. This meant finding ways to connect with our clients to better understand them; to connect the different parts of an organization, from R&D to customer services; to connect the leadership agenda to people’s daily work and vice-versa; and to connect core business processes. Data and AI should not be discussed only in the context of automation. If you think about what is required to make the typical R&D process successful, it’s the kind of people involved, and an understanding of the market and the clients. But typically, these processes are very disconnected. We need data to better understand the client and we need to use that data to get them the right people to work on a project, or identify if someone in the company has already solved a particular problem and might be able to provide helpful expertise. We reconnect the people and time horizons in organizations to help them avoid working in silos, and predict problems to lead success instead of firefighting. We predict the outcome of today’s actions or understand what happened six to 12 months ago that led to today’s situation.

Chivot: How has data-driven innovation impacted organizational culture and business models?

Syrjänen: We’re actually only at the start of that shift. If we exclude the largest companies, the impact of data and AI in traditional organizations has been rather limited so far. We’re not there yet but it’s going to be a major competitive factor in the future. The situation differs by sector. We see great progress and preparedness when working with media companies. They already use AI and data, employ data-driven business models, and use recommendation engines. At the other end of the spectrum, we work with the construction industry. Although this is a very traditional business, we see huge opportunities in an efficiency and business model revolution driven in part by data and AI. I’m an optimist, so within the next three to four years the impact will be felt more strongly, and we’re likely to see a significant acceleration of change.

One of the biggest challenges that we’ve identified is the significant gap between the business and the technical people—there is no proper dialogue. Companies often think transformational change is just about hiring more data people, but because their business leaders don’t understand what to do with that capacity, they’re only giving them small prototype tasks and don’t make full use of their potential. Whenever we discuss this topic with business leaders, our key agenda is to help them bridge that gap. Business people need to gain a better understanding of the opportunity data represents, and technical people need to understand the business side of things.

Training helps but it’s also about the overall maturity of an organization or sector. These things take time. Some industries, like banking, are more prone to digitalization, and the move towards a data-driven culture is smoother because what they have and work with is data. Others show a lot of interest but many are still missing the big picture. Business leaders can’t outsource transformation to their tech people. They need to become personally involved, which requires them to learn a whole lot, fast. This time requesting a 10-page brief is not good enough. You need to go deeper into what’s happening and what’s required, to gain a strong understanding of what change is going to be about and what it will and should look like.

Chivot: What are some of the other main obstacles to the adoption of data and AI by companies?

Syrjänen: First, of course, is data. Organizational data is frequently in bad shape. Many organizations want to create and have perfect data before they start doing anything with it—but the fact is you’re never going to have perfect data. Another challenge involves behavioral change. Technology may be difficult, understanding a business case may be difficult, but changing people is really the hardest part because changing people means changing their workflows, how they are used to doing things. What’s more, you need to convince them to adopt those changes. Most discussion around data and AI involves cool algorithms and new experiments, but those are not the real bottlenecks. Most of the technologies we use aren’t that fancy but rather quite standard and sometimes old. The challenge is connecting to business, reaching out to people, and then getting the data in a proper way.

Bringing business, people, and technology together is not only challenging, but also really inspiring. Often you’ll have two in shape, but the third element is missing. Maybe the technology and the business are there, but the people are forgotten; or the business and the people are there, but the technology or the data are missing. This is harder than traditional transformation because the concepts are more challenging. Our answer is to build conversations across these three domains. We want to help businesses gain a better understanding of technologies, offer them access to these tools, and allow them to experiment with them and implement them properly.

Chivot: How does Futurice help companies to leverage and connect these three aspects—business, people, and technology—in practice? 

Syrjänen: A client may have a pretty clear idea of the problem at stake and may just ask us to implement a technical solution. But what is interesting is to help them think further by bringing the people and the business aspects to the case. We then offer insights and the one that’s often most appreciated is the concept of proxy data. For example, projects in the construction industry tend to be unpredictable, and when predictions exist they often lack accuracy. A project’s challenges often are that budget or timetable overruns become visible only one to two months before that project is supposed to be ready, and when you figure that out, everything is already lost and it’s too late. We were once given the task of figuring out if we could use data to predict problems six to nine months before completion. We first tried to model the construction site through official and obvious data sources. The outcome wasn’t satisfactory because that data didn’t allow us to build a prediction that would give a heads up early enough. We then looked at a proxy indicator: Access control data, i.e., the movement of people through the physical gates of the construction site. It turned out that this very simple data actually predicted how the site’s construction project would fare in terms of budget and timeline.

If there’s a wicked problem, and you want to get an early warning, we look at what can be the proxy—and that can be many different things. This is actually something investors do. They look at harbor data, which reacts faster than GDP data, or satellite images from China to predict the evolution of the country’s GDP. Or they calculate oil inventories. This is a key part of what we bring to the table: You can solve a challenging problem by providing insights generated by proxies.

One of our key tasks is developing data infrastructures for our clients. We build and clean up data flows, process data more efficiently, and enable its fast optimization. How to present the data to people in a way that they can understand it and make judgement calls based on it is also important. There’s more at play than algorithm optimization.

We help companies address privacy concerns and data policies in an insightful way. For example, we developed the “Bubble Burster,” a search engine that tells you who knows about what, based on people’s public digital footprint within the company. We address privacy concerns in many ways—traditionally, e.g., through filtering, but also ingeniously, e.g., via transparency—which is one of our values. To make sure people are not searching for the wrong things, there is a user audit log. All searches are publicly accessible to all other colleagues. This transparency creates positive self-control.

Chivot: Can you give a couple of examples of concrete projects that show how Futurice solved problems with its clients?

Syrjänen: While our own transformation is related to applying data and AI to knowledge work, the help we offer our clients is, of course, broader. We helped Fortum, an energy company, take advantage of their data to optimize hydropower production. We worked on three areas related to the impact of the optimization model. First, by managing and storing input and output data in a structured way; second, through presentation to the end-user, including interactive editing of inputs and visualizing optimization results, using a domain-specific visual templating language; and third, by providing system performance optimization for all the links in the chain that call the optimization engine—from background systems to the browser and generated reports.

We created another energy-related solution with Elenia. It combines the main network and control system data with weather forecasts. The system uses data sources and historical data to predict and estimate possible damage to the network in a specific area. The estimate is displayed in Elenia’s control room and the partners’ resource management’s status map, on an hourly level, and it is regularly updated when a major disturbance progresses.

We’ve worked with companies in media and retail to build recommendation and personalization engines to create transformative user experiences.

We developed the “Medicine Radar,” surveying how patients use and discuss medications. This is also a proxy data question. If doctors ask their patients directly how they use the medication they’re provided, they may not hear the actual truth. The idea of the “Medicine Radar”, as we applied it, is to collect data from a popular website’s anonymous discussion forums on how people discuss healthcare and medication. We looked for information such as the dosage people actually take and the types of drugs they use to address certain symptoms, which can be compared to recommendations.

You can apply this logic of the symptom radar to an organization’s strategy, too. Executive leaders may build a strategy and then think their employees understand and apply it intuitively. What if we match data from actual daily work to strategy? We could gain visibility into whether there is a shared, proper understanding of it. It’s a great way to use raw data to build a holistic understanding of what really happens in practice, and to help an organization’s strategic and cultural change. Whenever there is too much distance between people who need information and people who provide that information, then all we’re making are assumptions. With data, we want to get rid of assumptions and see what really happens.

Together with a large Finnish newspaper, we have applied this approach to generating data during the coronavirus crisis by collecting voluntary symptom reports from the general public, to raise awareness of where COVID-19 symptoms are concentrated. The objective was to help identify where testing capacity could be too limited, depending on the geographical spread of the disease.

We created a facial recognition system for hands-free check-in, which we trialed at Helsinki Airport with Finnair and Finavia, using (untraceable) biometric data. The current crisis and resulting restrictions on physical contact make it a potentially very useful tool. Passengers can scan their face with an app on their phone and use that to register for check-in. When making their way to the check-in desk, they would then already be identified and their boarding passes, already printed—without them having to take out their travel documents. We used the same technology for a facial payments system piloted in Oslo by TINE Group.

We also ran a pilot of the Bubble Burster for a healthcare provider. When someone contacts a call center to reach out to a healthcare provider, the current process of how a doctor is assigned to the patient is very inefficient, and based on rudimentary formats such as simple texts. We used the digital footprint of doctors to build a system that assists the person you’re calling to easily figure out who to recommend. Based on the symptoms you describe, he or she can enter keywords that point to the right specialized doctors.

You may also like

Show Buttons
Hide Buttons