In this interview, Marko discusses how data can help identify key operational improvement opportunities, why companies should approach AI pragmatically and how businesses can get started without large-scale system projects.
Hi Marko! Could you briefly tell us about your background and how you ended up at Mantec?
Hi! I have spent practically my entire career working on projects. Before joining Mantec, I worked in international development projects and travelled extensively around the world. Some of the most memorable experiences came from working with street children in Northern Mozambique and establishing HIV-related projects in Central Mozambique. Those experiences gave me a valuable perspective on persistent project work. The challenges were enormous, yet at the same time no problem felt impossible to solve. Compared to that, development projects in companies and organisations take on very different proportions — the problems are usually ultimately quite limited, the goals are clear and the obstacles manageable.
After completing my studies, I started looking for the next step in my career and ended up in management consulting. Quite soon after, I was headhunted to Mantec, where I have spent most of my consulting career.
Academically, my background is in analytical philosophy, with a long minor in economics. My studies also included business-related subjects and exchange studies at the University of Michigan. My main specialisation was business ethics, particularly business ethics in competitive environments.
Analytical philosophy is actually a very practical background. At its core, it is training in thinking: learning to build logical chains of reasoning and critically examine the assumptions behind them. You also learn to distinguish valid arguments from flawed ones. This has been extremely useful later in both IT roles and consulting work — in practice, the core of my profession is exactly that: questioning assumptions and separating real cause-and-effect relationships from apparent ones.
In client work, my philosophical background is visible for example when analysing processes, evaluating chains of impact, or assessing investment decisions by seeking comprehensive arguments both for and against different options and understanding the scale of different effects. A concrete small-scale example comes from the textile industry, where we conducted a waste analysis by product and product type. Some models had significantly higher waste rates than others, which initially made them appear to be poor-performing designs. When we analysed the issue more closely, however, we discovered that defects were more common only in specific parts of the textiles used in those models. Eventually, we found that because the patterns were largely symmetrical, they did not always work properly in certain textile sections due to the thread twist direction affecting how the textiles behaved when stretched. To reduce waste, safety margins in cutting needed to be adjusted according to the thread direction. Data analytics helps identify where deeper investigation is needed and allows us to drill down as far as necessary, while analytical thinking helps uncover the real causes among the apparent ones. After that, the key question becomes scalability: how do we provide people working permanently in the organisation with the tools to identify these problems in everyday work, analyse them, solve them conceptually and ensure the solutions are actually implemented?
I try to choose projects I can genuinely be proud of. For example, we have worked on fantastic projects within the Finnish media sector, helping ensure literature remains vibrant in Finland and continues to have strong distribution channels. In healthcare, we have supported projects that ensure and expand the capacity to produce life-saving medicines. We have also contributed to strengthening Finnish resilience against external threats, both in terms of information flow and national defence. These are all projects I have been very happy to work on.
I personally aim to act ethically and also help organisations navigate similar questions in their own operations. Companies operate in highly competitive environments and ethical conduct affects everything from competitiveness and employee satisfaction to customer satisfaction, risk management and costs. Competitors’ actions in the same markets inevitably influence a company’s own position and I strive to help my clients maintain the strongest possible competitive position even when making bold decisions aligned with their principles. In many projects, we are faced with choices between alternative solutions — and my recommendations are always ones that can stand public scrutiny.
I started my consulting career over 20 years ago. Along the way, I have also worked with international IT teams and deepened my expertise in areas such as computer science and statistics. Personally, it has always been important for me to stay close to the edge of new knowledge and development. In client organisations, you can clearly see that those who keep learning throughout their careers grow much faster and further than those who stay tied only to their original education. This applies equally to our clients and to consultants like us.
You specialise in data analytics and the utilisation of AI — what does this mean in practice?
Most of my work revolves around client projects. For example, I conduct assessments related to the use of data and participate in Mantec’s projects by building practical solutions tailored to clients’ needs. I develop various data-driven management and monitoring models for companies — essentially ways to run businesses more effectively using information.
Over time, I have gradually become someone who is especially involved in projects with a strong data or analytics angle. My role varies: sometimes I support other consultants and at other times I am directly involved in analysis work and implementation.
What I find particularly interesting is quickly understanding a company’s situation through data — seeing what the business looks like from an analytical perspective and identifying the key factors that explain performance. This combines both business understanding and analytical capability.
I am also involved in projects where clients face challenges related to data management or need to develop their capabilities further. In those cases, my role also includes implementing practical solutions and helping integrate them into everyday operations.
In addition to client work, I also contribute to internal development at Mantec. For example, I have focused extensively on AI and on how we at Mantec can strengthen our own capabilities and utilize new technologies in client work. In a way, I also consult our own organization on these topics. I held our first internal AI trainings around eight years ago in the context of machine learning, and since then I have continued to deepen my own expertise while regularly training our consultants on developments in the field several times a year.b
We continuously evaluate where technologies have matured enough to create real value either for us or for our clients — in other words, we try to separate hype from genuine progress. At the same time, we constantly consider which client challenges could actually benefit from a specific technology. This is an ongoing discussion because the answer evolves together with technological development.
My background in business ethics has proven surprisingly useful specifically in AI and data-related questions. In business ethics, you constantly need to distinguish between what is legal, what complies with contracts, what is defensible from the customer’s perspective, and what is sustainable in the long term. The same type of reasoning applies when considering what data can be fed into AI systems, which systems can be approved for use, and under what conditions. In AI, the pace of technological development is currently much faster than the pace at which companies’ contractual practices and internal policies evolve.
In what kinds of situations and use cases do clients typically rely on your expertise?
The use cases vary significantly depending on the client, but in many cases the work focuses on optimizing business operations through data. This can involve reallocating capacity — understanding what is actually being done, how much of different products can realistically be sold, and how to ensure production lines deliver what has been promised to customers. It can also involve waste analysis — identifying how much production waste or other inefficiencies are generated and systematically reducing them. Logistics improvements are another common area: batch sizes, routes, services, packaging methods, storage solutions, shipping methods, inspection practices, and so on. Based on these findings, we can implement concrete operational changes and improvements. These are not exotic topics, but every industry has its own specific characteristics and opportunities on top of them.
One common pain point for clients is time horizon. A business need emerges, but when it is handed over to a technical unit that may not fully understand the business context, the outcome is often a six- or twelve-month implementation plan that only partially addresses the original need. In our projects, timelines are typically measured in months rather than years, so we often tell clients: this can be implemented in two months instead of two years. In practice, I frequently translate the client’s business need directly into a technical solution without unnecessary intermediaries.
A key principle in my work is understanding what information is relevant to different roles within the organization. What matters to senior management is often very different from what matters to middle management, supervisors, or employees on the operational level. Based on this, we build customized views and tools for each level that support their daily work: helping them monitor situations, provide feedback, plan ahead, and make better decisions in everyday operations.
An important part of my role is acting as a bridge between business and technology. Clients may have a very clear need, but transforming it into a practical solution can be slow in many organizations. This is where I can often create value: helping identify solutions that achieve the same end result faster and with less complexity.
AI is currently on the agenda for many companies — what are the most common challenges when trying to turn it into concrete business value?
At Mantec, we have followed the development of different forms of machine learning for well over a decade. The biggest change during these years has been the rise of generative language models into public awareness. Machine learning technology itself was not new, but now it has become accessible to practically everyone. Along with that, expectations — and sometimes disappointments — have also grown.
Although the market talks a lot about massive productivity leaps, the reality is more moderate. That is why my own approach is very pragmatic: I continuously monitor where different technologies are in terms of maturity and at what point they begin to create genuine value for clients. As new, truly useful applications emerge — whether in computer vision, process automation, or elsewhere — we adopt them and integrate them into client projects.
Regarding productivity, I would put it this way: the dramatic productivity revolution promised in articles has not really materialized anywhere yet. The actual benefits are more realistic and emerge through individual tools when implemented properly. Personally, I use AI models extensively in coding. In my work, I constantly need to learn new environments quickly — one day it might involve Excel, Power BI, VBA, and DAX; another day SAP macros, Python, SQL, and different web platforms. I also use AI in analyzing unstructured data, retrieving public information, and finding relevant articles. In addition, I have developed some semi-standardized workflows where AI agents perform certain tasks on request. Similar incremental benefits are gradually emerging for clients as well. However, large-scale transformations that completely reshape businesses are still developing in most industries.
New opportunities are beginning to emerge around AI-based processes and service chains, but in many areas the technology is not yet fully mature. A good example is the chatbot functionality on many corporate websites — they have improved, but their practical performance still often falls short of expectations. At the moment, the most significant benefits are tied to specific tools such as Copilot-type solutions.
One major topic around AI is cybersecurity and the risks associated with it. In particular, many companies are uncertain about what information can be shared and where data can be entered. Because of this uncertainty — or because the perceived risks feel too high — companies may restrict AI usage quite heavily. However, many organizations already have data spread across numerous online services: SharePoint, email, Teams, dashboards, intranets, or equivalent platforms. In many cases, user permissions and access rights for sharing that information have already been defined. One practical step toward AI adoption is to use these existing access structures as the foundation for AI solutions as well.
Another challenge is how much freedom employees are given to take initiative. Are employees allowed to automate their own work using code, macros, Zapier, Power Automate, or AI agents — or are all of these tools locked down? In many development projects, we help companies find the right balance: maintaining cybersecurity while enabling people to learn, take initiative, and develop their capabilities with as little friction as possible.
We discuss these topics extensively internally as well: which tools we can use and under what conditions. Often, boundaries also need to be set — some data simply cannot be used or transferred to external systems. The same applies to our clients. It is important to understand what is acceptable and secure from their perspective and act accordingly.
“Operational data is an untapped goldmine for many companies”
What does effective data analytics look like in everyday operational work?
The starting point is that companies almost always have far more data than they are capable of utilizing. In the past, organizations might have had one hundred times more data than they actually used — today, the amount may be a thousand times greater or more. This is not accidental. Financial data must be analyzed because the law requires it — companies have finance departments, controllers, and accountants, so financial information is well managed. But operational data, which is collected in vastly larger quantities, is not analyzed with the same discipline by anyone. That is why it is an untapped goldmine for many organizations.
The solution is usually not to collect even more data or introduce new systems. More often, the real opportunity lies in extracting greater value from the data the company already has. Data collection is frequently at a good level, but analysis and utilization are where the greatest development potential exists. That is where the large, often untapped opportunity lies.
My starting point is usually to make the most of the tools the client already has in place. Many of our clients operate heavily within the Microsoft ecosystem — using Office, SharePoint, Teams, and increasingly also Power BI. Recently, we have built many solutions around Power BI, such as management views and dashboards that turn raw data into usable insights for the right people.
At the same time, Excel still plays an important role. Although some organizations want to move away from it, in practice it remains an extremely flexible tool for analytics — especially for rapid ad hoc analysis. Excel also allows companies to pilot new solutions quickly: testing whether a certain view or analysis is genuinely useful and actually used. If the pilot proves valuable, the solution can later be integrated into a more permanent system. Modern Excel’s data modeling capabilities are essentially on the same level as many dashboard applications. If a dashboard does not need to be shared across a large organization, Excel can actually be one of the best solutions available. It can retrieve data from databases, the web, ERP systems through macros, and local files; combine tables; enrich data; and ultimately visualize it more flexibly than many other tools.
In addition to these tools, we use code-based solutions when necessary, but in practice they are more often tools for research and deeper analysis rather than permanent systems delivered to clients. Microsoft technologies alone can take companies surprisingly far, and there is usually no need to build heavier systems before it has been clearly proven that lighter alternatives are insufficient. It also makes little sense to leave behind solutions that the client organization cannot maintain or further develop independently. Operational data analysis is part of a company’s core competence and should absolutely not be outsourced entirely. In many of our projects, developing this competence is itself a core part of our work.
How can companies start utilizing AI in a practical way without making things overly complicated?
One key theme is access to information and tools. To use data and AI effectively, people need sufficiently broad access to the tools the organization has chosen — naturally while taking cybersecurity into account. The company’s role is to determine which tools are secure and appropriate. After that, access rights should generally be distributed as broadly as possible so that employees can begin applying the tools in their own work.
Because everyone’s work is different, the best use cases often emerge organically in daily operations. When people have access to tools and data, they can discover more efficient ways of working themselves. The next step is to share those learnings internally — making good practices and successful solutions visible so that others can benefit from them as well.
I’ll say this directly: in many organizations, the biggest bottleneck is not technology or model quality, but the fact that companies withhold information from their own employees. Access rights are tightly restricted, and employees are expected to know in advance what data they need and request separate access to it. Naturally, people cannot make use of information they do not even know exists. This can often be solved without new systems or new budgets — it requires a decision that information should, by default, be open, with only limited areas restricted.
“Openness and transparency in data availability are fundamental prerequisites for effective utilization.”
On a practical level, progress begins by giving people the opportunity to learn and experiment themselves. If a company wants to move forward with data and AI, the first step is to consciously choose which tools will be used. For many organizations, a natural starting point could be something like Microsoft Copilot, particularly the Enterprise version where cybersecurity has been taken into account and which integrates with existing data environments such as SharePoint and Teams.
Once the tools are introduced and made broadly available, the next step is to guide employees in using them — for example by teaching them how to build simple AI agents or utilize chat-based functionalities in their own work. Through this, the organization gradually develops: people learn to use new technologies in their daily work, and the company simultaneously builds the capability to utilize them more broadly as the technologies mature further. There is one important lesson here: companies should wait long enough for technologies to become mature enough to be useful, but not so long that all competitors have already implemented them and built internal capabilities before they even begin.
Can you share an example where better use of data or AI improved a client’s operational performance?
One of our industrial manufacturing clients had significant variance in production performance. Each production process contained dozens of stages, there were hundreds of different products, and the phase times stored in the ERP system were based more on rough estimates than on real data.
We built a data model where each process stage could be compared against all previous runs of the same product to identify whether a particular stage represented a problem, normal performance, or better- or worse-than-usual execution. After that, we were able to predict the completion time of each stage and build production control mechanisms that allowed supervisors to prepare proactively for the start and completion of different phases. At the same time, identifying deviations enabled stage-specific problem-solving from both technical and product quality perspectives.
This was only one development initiative among many, but overall the company’s production volumes increased by more than 30% compared to the year before the project. Perhaps even more importantly, this type of analytical approach creates the foundation for continuous improvement: every future production batch and process stage can also be compared against this improved performance level, making it possible to see which direction performance is heading. The success of the whole ultimately comes from the success of its individual parts.
Data-driven management is at the core of what we do and strongly present in all our client projects. One of our key principles is that frontline management plays a critical role in business performance — that is where day-to-day operational work actually happens. That is why we strive to bring relevant information specifically to supervisors and frontline managers and provide them with views and data they may never have had access to before. When frontline management has better visibility into operations, they also gain the ability to improve them in practice — where the real work takes place.
In many cases, we begin by assessing together with the client what kind of development potential exists within their operations. After that, the key is ensuring that the identified goals are actually achieved.
There is almost always a strong mathematical element involved in this work. All development actions must somehow be quantified — meaning their impact must be understood numerically — so that they can be systematically managed. In practice, this means continuously measuring progress and steering operations toward the desired outcomes. Without measurement and concrete metrics, improvements are difficult to implement or sustain.
How do data and AI connect more broadly to operational development and performance improvement?
There is a common saying that “you get what you measure,” and that is very true. If something is not measured, it becomes extremely difficult to know whether progress is being made toward a goal. And if you do not know that, it is also difficult to understand which factors are influencing the outcome.
In practice, this means first defining a clear target — for example a production volume within a certain time period. After that, the key factors influencing that goal are identified and systematically measured.
In manufacturing, for example, this could involve analyzing equipment downtime, its root causes, and which of those causes can actually be influenced. Once you know which factors are controllable, you can implement concrete improvement actions and monitor whether they are producing the desired results.
This leads to a broader system where analytical models, reporting structures, and forecasts are continuously built: defining targets, understanding the factors influencing them, predicting future developments, and steering operations in the right direction. Measurement, forecasting, and operational control are all central parts of how companies are practically managed through data.
I have repeatedly seen internal company projects fail in a few predictable ways. First, companies assume they need more data when in reality the solution is usually better use of existing data. Second, they measure what is currently being discussed rather than what actually drives operations. Third, they focus too heavily on financial data and too little on operational data. Fourth, they build dashboards without agreeing on who will actually use them and when — which means the dashboards never influence decision-making.
There is no single universal answer to how AI should be used in operational development. Everything depends on what is relevant for each specific company. The key is identifying whether a certain aspect of AI or machine learning is genuinely useful in that particular business environment. In many cases, the application revolves around detecting deviations or anomalies.
In the financial sector, for example, AI is used for fraud detection — identifying situations where customer behavior deviates from normal patterns. Similar use cases exist in other industries: in manufacturing, AI can detect when equipment behavior differs from normal, or identify unusual fluctuations in operational processes. These findings often serve as triggers for deeper analysis and corrective action.
The important thing to understand is that there is no ready-made “one solution fits all” AI model. No single AI solution will automatically create major improvements for every company. That is why the work is always case-specific: you must identify the situations and triggers where AI genuinely creates value and build solutions around those.
If a company wants to move forward with data and AI, where should it begin in practice?
Many companies are currently wondering whether they should first invest in new systems, adopt AI tools, or build an entirely new data strategy. Often, however, the best starting point is much more practical: first understand what data the company already has, how it is currently used, and where in daily operations it could create the greatest value.
In practice, we typically approach this through three steps: first, (1) a short company analysis; second, (2) mapping information needs across different management levels; and third, (3) selecting tools and supporting implementation.
At Mantec, client work usually starts with relatively low-threshold engagement. A typical starting point is a short company analysis, during which we spend a few weeks examining the client’s operational activities, identifying development potential, and recognizing the most important impact points. For the client, this represents low risk, but almost without exception the analysis itself already creates value simply by increasing understanding of what is happening in the business and where the real improvement opportunities lie.
I honestly cannot remember participating in a company analysis that did not pay for itself through the findings alone — even without launching a full development project afterward. If a company wants to explore these topics with us, that short company analysis is usually where we begin. It covers both data-driven management and broader operational development opportunities, while also producing a roadmap for how to move forward — either together with us or independently.
Another practical way to begin is by asking people across different organizational levels what information they actually need in their work. In workshops, for example, you can discuss what information is critical for successful performance, whether it is available, and whether it arrives in the right format and at the right time. This often reveals the current state very concretely: some employees have built complex personal systems to combine information from different sources, while others have very little data available to support decision-making. Even these discussions alone can reveal significant development opportunities.
The approach to AI is similar. First, organizations select the tools they want to use — taking cybersecurity into account — and then make them broadly available to employees. At the same time, implementation is supported through training and practical examples.
One particularly important perspective in AI utilization is process analysis. Instead of thinking about entire jobs, companies should identify specific parts of processes that can be automated or improved with AI support. Most roles are multifaceted, but within them there are repetitive tasks where AI can create real value. In these cases, a certain step can be fully automated, or the employee can receive a ready-made suggestion that they can approve, modify, or reject. This makes AI a practical everyday tool that supports work instead of replacing entire roles.
If there is one thing to take away from this discussion, it can perhaps be summarized into two questions for the next management team meeting:
First: who in our organization does not currently have access to the data they would need for better decision-making?
Second: which specific process step — not role — would be the best candidate for our first AI experiment?
More often than not, these two questions lead to a far more useful development path than any new system investment.
