An IT manager at a mid-sized manufacturing company recently put it plainly: ‘We want to start a machine learning project, but we don’t even know where our data sits.’ That single sentence captures the contradiction at the heart of many Turkish businesses right now. Artificial intelligence and big data have moved from technology conferences into boardroom agendas, but the excitement often outpaces the groundwork. Launching an AI initiative without a data strategy is, in practice, like trying to build the upper floors of a structure before laying the foundation.
A data strategy is the structural framework that defines what data a company collects, for what purpose, how it is stored, how its quality is assured, and who can access it under what conditions. AI models sit on top of this framework; they are not the foundation itself. No matter how sophisticated a machine learning algorithm is, if the data feeding it is dirty, incomplete, or inconsistent, the output will be unreliable. The principle widely known in data science as ‘garbage in, garbage out’ applies with full force to AI investments as well.
Where do Turkish companies actually stand on data? The mandatory adoption of e-Invoice and e-Ledger systems has significantly increased the volume of structured digital data that businesses generate. ERP systems, CRM tools, and cloud-based applications have reached more companies each year. But the presence of these tools does not mean a company has a data strategy. In most businesses, financial data lives in the ERP, customer data sits in a separate CRM, and the sales team’s notes are scattered across Excel files and email inboxes. These data islands do not communicate with each other, making integrated analysis — and therefore any meaningful AI application — nearly impossible to execute reliably.
A workable data strategy rests on three core components: inventory, quality, and access architecture. A data inventory is a catalogue that documents all data sources in the company, their formats, update frequency, and the business unit responsible for each. A quality program consists of the processes that regularly measure and improve data accuracy, completeness, and consistency. Access architecture defines which users can reach which data under which conditions — critical for both operational efficiency and data security. An AI infrastructure built without these three components in place will either become unusable within a short time or produce decisions that mislead rather than guide.
Consider a concrete scenario: a retail company wants to use machine learning to predict customer behaviour. To do so, it needs historical sales data, customer segmentation records, and inventory movements. But sales data is spread across three separate systems, and the customer codes in those systems do not match each other. Inventory data is updated manually each week and frequently contains missing entries. A predictive model built on this foundation will perform far below expectations in real-world conditions. When total cost of ownership (TCO) is calculated honestly, the time and money spent on data cleaning and integration typically exceeds the investment in the model itself.
The biggest barrier to building a data strategy is not technical — it is organisational. The question of which business unit owns which data rarely has a clear answer. Disputes over data ownership between finance, IT, and operations slow projects down considerably. Data quality improvement also creates short-term workload, which makes middle managers inclined to defer it. Senior leadership must treat data strategy not as an IT project but as a company-wide priority. Without that top-level commitment, the organisational resistance is difficult to overcome, and AI pilots stall before they produce any measurable ROI.
If you are a decision-maker evaluating an AI investment, start by answering three questions honestly: What specific business problem are you trying to solve, and what data does solving it require? Where does that data currently live, in what format, and how often is it updated? Who monitors data quality, and through what process? If you cannot give clear answers to these questions, directing your AI budget toward data infrastructure first will deliver significantly higher returns over time. Artificial intelligence is not a destination in itself — it is a tool that creates genuine value only when built on a solid data strategy.
This article was originally written in Turkish by Gökhan MERCANOĞLU on June 6, 2016 and has been automatically translated into English and other languages using machine translation.