AI Development
April 23, 2025
11 min read
In today’s world, data is more than just a byproduct of business operations. It’s a strategic asset. A lot of organizations invest in data tools and technologies. Nevertheless, the real challenge lies in creating a culture where data is understood, trusted, and consistently used to drive decisions at every level. How to do it? That’s one of the key questions that Max Golikov, the Innovantage podcast host and CBDO at Sigli, discussed with his guest Dr. Carsten Bange.
Dr. Bange is the founder and CEO of BARC and an expert market analyst for data analytics and AI. For over 25 years, he has focused on evaluating software vendors and technologies, helping organizations make informed decisions based on market trends, strengths, and weaknesses of various solutions.
While BARC began as a technology advisory firm, its scope quickly expanded. Today, the company also supports clients with data and AI strategy, organizational development, and data governance. All this is a key foundation for successful analytics initiatives.
A major area of Dr. Bange’s work is data culture, which is the human side of data-driven transformation. As he emphasized, even the most advanced technologies can fail without the right mindset, skills, and engagement from people. Lack of adoption is often not a technical issue, but a cultural one.
To promote this idea, he launched The Data Culture Podcast. He interviews practitioners and leaders who share their experiences in building a strong data culture within their organizations.
As Carsten explained, data culture refers to the unwritten rules that shape how organizations work with data. It encompasses the values, beliefs, and behaviors that support the effective and ethical use of data across the organization.
At its core, data culture defines how people think about and interact with data, how they use it, and for what purposes. Besides that, it determines how organizations leverage data to drive decision-making, process improvement, and innovation.
One of the most common questions organizations face is: “How to build a strong data culture?”.
Dr. Bange highlighted that “data culture eats data strategy for breakfast” as even the best data strategies fail without the right behaviors and ways of thinking.
Implementing data culture is not about issuing a directive. Culture can’t be turned on or off. It must be influenced through ongoing, targeted efforts.
Carsten shared his framework that organizations need to tackle when they want to improve data culture.
This framework includes 6 areas.
Enhancing data literacy is often the starting point. Upskilling employees, increasing their confidence and competence with data, as well as fostering a shared understanding of its value are foundational steps.
Data culture depends on data access. In many organizations, it is limited either due to technical constraints or restrictive access rights.
There are two models of data access. The first one is “need-to-know”. It presupposes that access must be requested and approved. The second one is “right-to-know”. It lets data be open by default unless it’s sensitive, like HR or personal information. The latter fosters trust, openness, and initiative. People have access to data and they can use it to bring benefits to their organizations.
According to Carsten, communication plays a vital role in reaching people and shaping their behavior around data. To build a strong data culture, leadership must consistently communicate the strategic value of data to their employees. They should show how data aligns with business goals and supports competitive advantage. It’s also worth sharing success stories and role models. Real examples of how data has driven results, like increasing revenue or gaining new customers, can motivate others to change their attitude toward the data they have.
A successful data strategy must be based on the existing data culture. Ambitious plans for enterprise-wide AI or advanced analytics are unrealistic if employees lack the tools, skills, or access to data.
Too often, strategies are overly technical. They are focused on architecture or infrastructure. But they neglect the people who must use those tools. Data culture should be an integral part of any data strategy to ensure alignment with organizational reality and to support real execution.
Strong leadership is critical to fostering a data-driven culture. While grassroots efforts are valuable, they reach a limit without top-down support. Senior leaders must actively promote data initiatives and model the behaviors they want to see.
Carsten pointed out that the biggest blockers are often in middle management, where key resource and access decisions are made. If middle managers withhold support, it can slow down cultural progress.
This component is about balance. Too little governance leads to chaos. Too much creates fear and resistance. Overly strict rules or legal-heavy processes can discourage people from working with data at all.
Effective data governance should enable data use, not restrict it. It should guide employees, support data quality, and create clarity without driving anxiety. In a positive data culture, governance is seen as a help, not a hurdle.
According to Carsten, company size is not the deciding factor when it comes to benefiting from a strong data culture. That’s a conclusion that he has made after years of working with a wide range of organizations and interviewing nearly 150 guests on The Data Culture Podcast.
Of course, large organizations often have more resources to work with data. For example, they can form dedicated data culture teams. Such teams may focus solely on promoting data literacy, leading internal communication efforts, and organizing events like annual award ceremonies celebrating successful data projects.
This structured approach allows them to scale data culture initiatives across the enterprise.
Smaller companies may not have formal teams but they can still adopt the same principles. While the scale and execution differ, the core concepts and framework remain fully applicable.
In his discussion with Max, Carsten mentioned a strong link between overall company culture and the success of data culture initiatives. For example, organizations moving toward data products tend to succeed when their company culture already promotes collaboration, openness, and knowledge sharing. In contrast, organizations with siloed, disjoined cultures often struggle with such approaches.
Among forerunners in a data culture, Carsten named Merck, a global pharmaceutical leader, that has a dedicated data-focused team.
Before the first pan-European Data Culture Summit, Bange conducted a study to identify organizations actively investing in data culture roles. The research, based on LinkedIn data, revealed that:
The financial services sector, including banking and insurance, is currently the most active in data culture, accounting for over half of all identified roles. This is likely due to the industry’s data-heavy nature and strong regulatory requirements for data governance and quality.
Effective data governance must align with an organization’s structure and operational reality. "One-size-fits-all" models don’t work. Your model must be adapted based on whether the company is centralized or decentralized.
Many organizations today are moving toward decentralization of structure and data ownership. This reflects a long-term trend in data and analytics: shifting responsibility and ownership closer to business units. This is also often accompanied by decentralizing platforms, tools, and access.
Such a shift challenges traditional ideas of centralizing all data in one place. The once-dominant data warehouse approache, which aimed to consolidate all data centrally, is no longer practical for many organizations. The growth in data volumes, the rise of real-time IoT data, and increasing complexity make it difficult (and sometimes even impossible) to bring all data together in a single location.
Instead, modern data architectures often follow distributed models, such as data fabric, which help to maintain a coherent framework for interoperability and governance.
According to Dr. Bange, the engine behind decentralization in data and analytics is the need to scale data usage across the organization. To create a strong data culture, companies need to empower more people to actively work with data and analytics tools. Centralized models often create bottlenecks either in data access or in the limited availability of central data teams.
In many cases, central data teams are overwhelmed and can’t fully support the growing demand for analytics. As a result, business units need to decide whether they should wait or take the initiative themselves.
Decentralization becomes a logical step here. Thanks to this approach, teams can access and integrate their own data, build local data capabilities, and act autonomously.
One major benefit of decentralization is proximity to domain knowledge. Domain expertise is critical for building meaningful analytics or AI models. Being closer to the actual business processes allows teams to identify relevant use cases, involve stakeholders early, and ensure real-world adoption.
This is especially important when transitioning from pilot AI projects to enterprise-scale deployment. The main challenges at this stage are often organizational, not technical. Scaling AI and analytics requires changing workflows and embedding new tools into existing processes. All these issues can be addressed faster when data teams are integrated within the business units.
However, entire decentralization is often not the best choice. Here is when hybrid models come into play.
Hybrid models offer the most practical and scalable path forward for organizations navigating data governance. Dr. Bange explained that this approach strikes a balance between central oversight and decentralized autonomy. It means that it can adapt to organizational complexity while enabling growth.
There are two key reasons to centralize certain aspects of governance. First of all, some topics are too critical to leave to individual teams. Regulatory compliance, such as GDPR, is a prime example. Instead of having dozens of teams interpret and apply these rules independently, centralized governance ensures consistency and reduces risk.
Secondly, limited expertise in emerging areas like AI often requires a centralized starting point. Over time, as capabilities mature, these roles can be gradually decentralized and central units are shifted to supporting roles, like education and community-building.
At the same time, organizational diversity plays a crucial role. Within the same enterprise, different departments or regions can be at vastly different levels of data maturity. Some may have strong internal teams, platforms, and domain expertise. Others may rely heavily on centralized support and shared services.
A hybrid approach acknowledges such differences. It allows flexible service models, where units can choose what they handle independently and what they consume from central teams.
The rise of AI has significantly shifted the conversation around data in organizations. What was once a specialized concern for data teams has now reached the boardroom. Executive leadership increasingly recognizes that AI requires high-quality, well-governed data to deliver real value.
This understanding has reinforced the need for robust data governance practices. As companies aim to expand their AI capabilities, they must also address long-standing challenges around data quality and accessibility.
The roles of data and AI literacy are equally important across the organization. Just as with broader data culture efforts, successful AI adoption requires behavioral and mindset shifts. Employees must understand what AI is, how to use it, and feel empowered to experiment with it.
Access not only to quality data but also to AI tools and infrastructure remains crucial. Making AI capabilities widely available within the organization democratizes innovation but also increases the importance of governance frameworks to guide ethical and compliant usage.
The introduction of the European AI Act underscores this point. While some organizations view it as restrictive, others see its value in providing clarity. With it, companies have received a stable framework within which they can build and scale
When it comes to becoming data-driven, organizations face a common dilemma: should they fully rely on large language models and hope that AI is smart enough to help them work with data, or should they take a more conservative route and focus first on cleaning and organizing their data?
Dr. Bange believes the real challenge is doing both at the same time. AI often acts as a trigger for companies to finally take a closer look at their data. Poor quality, outdated models, and years of underinvestment in data infrastructure are typical issues.
Ideally, organizations would fix their data first and then build AI use cases on top. But that approach isn’t realistic in a fast-moving environment. Nobody wants to hear that leveraging AI requires two years and several million euros just to clean the data.
According to Carsten, it could be sensible to opt for a more pragmatic approach: find use cases where AI can deliver early value while simultaneously improving the data foundation. Such projects can demonstrate the potential of AI. They also provide time to make the necessary long-term investments in data quality.
There are two major blind spots for organizations trying to implement data culture.
The first one is the human element. Amid the excitement around new AI models and technological advancements, companies often don’t notice the central role of people. As AI automates more tasks, the need for human oversight and engagement becomes even more critical. Building a strong data culture isn’t just about tools. It is also about collaboration, and continuous learning.
The second blind spot is underestimating the speed of technological change. Many organizations lack a clear grasp of how rapidly AI is evolving. This can make them slow to adapt or experiment. As a result, they may fall behind their more agile competitors that embrace AI-driven automation and innovation more quickly.
At the end of their talk, Max asked Carsten to share recommendations on how to start building a data culture at an organization.
The first tip was quite simple: just to start. Too many organizations hesitate or overthink the process. However, taking action is vital.
He also recommended using a structured framework, such as his own model with six key areas that influence data culture. This framework helps organizations assess where they currently stand and identify which aspects need the most attention.
Dr. Bange also mentioned two areas that are often underestimated at the beginning of the journey: data access and data communication. Many companies don’t realize their importance until they are already a year or two into the process. And this can become a serious obstacle for them.
Want to get more expert insights into how to boost your business growth in the data-driven world? New Innovantage podcast episodes will shed light on this! Stay tuned!
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.svg)
.png)
.png)
