Do You Know Which Mode You Are In When You Make Decisions?

Consider the general manager of a construction firm based in Kayseri — a mid-sized operation with 284 employees delivering roughly 12 projects a year. Last October, the company’s project analysis system flashed a green light on a mixed-use development in Ankara: demand indicators were solid, the cost model was internally consistent, and the cash flow projection cleared the threshold. The general manager still did not sign. ‘Something inside me said no,’ he told me. Two months later, a competitor announced a nearly identical project on the same corridor — they had moved first. The general manager himself shared this story, and he still could not answer his own question: ‘Did my instinct save me, or did it cost me?’The debate between data and intuition in decision-making has occupied management literature for decades. But in 2024 the frame around that debate has shifted fundamentally. Many organisations now have decision-support dashboards, AI-assisted demand forecasting, and real-time reporting layers. Access to data is rarely the bottleneck. The real problem is this: the majority of managers use data to confirm a decision, not to reach one. The decision has already been made — through intuition, first impression, or simple inertia — and the data is positioned behind it afterward to provide cover. This is not a data literacy problem. It is a decision-awareness problem. And the failure to distinguish between the two quietly undermines both the return on data investments and the capacity for organisational learning.My thesis is direct: balancing data and intuition is not the manager’s primary task. The primary task is knowing which mode you are in when you decide — and recording it. If you are deciding by intuition, say so explicitly. If you are deciding by data, understand what the model actually measures, and what assumptions are embedded inside it. Without a record, there is no accountability. Without accountability, organisational learning cannot begin. This framing is the only practical exit from the ‘data versus intuition’ trap that keeps recurring in management conversations.Intuition carries an unfair reputation — especially in organisations where data scientists hold significant influence. In practice, intuition is the brain’s accumulated capacity for pattern recognition built over years of exposure. The Kayseri general manager may have sensed that a competitor was acquiring land in the area, or noticed a subtle shift in the posture of a local subcontractor — signals that had never entered a CRM field or a project dashboard. The problem is not using intuition. The problem is using it without traceability, without the ability to examine whether it has been reliable in the past, and without opening it to challenge. When a manager says ‘something inside me said no,’ at least two questions should be answerable: Has this instinct been right before under similar conditions? And what triggered it this time? In most Turkish SMEs that institutional memory does not exist — and that is precisely where value leaks. In cyclical, relationship-intensive sectors like construction, tourism, and logistics, accumulated experience is a genuine asset. But it only converts into organisational value when it becomes visible and examinable.Data has its own blind spots, and they are just as costly. The most common trap for data-trusting managers is skipping the question of what the model represents rather than merely what it measures. A demand forecasting model learns from historical behaviour. It cannot learn about market discontinuities, the entry of a well-capitalised new competitor, or a silent shift in consumer preference — at least not until those changes have accumulated enough data trace to register. In Turkey’s construction and retail sectors throughout 2023, several firms read the temporary demand spike caused by inflation-driven purchasing behaviour as a durable growth signal; their models supported that reading. Demand forecasts that did not systematically incorporate sensitivity to foreign exchange rates and imported input costs told managers to proceed — and some of those managers faced difficult quarters as conditions shifted. The data was not wrong. The context was missing. This is precisely the most frequent failure mode of AI-assisted decision-support tools as of 2024: an inability to communicate to the manager the model’s confidence interval, the age of its training data, and the external shocks it has not been designed to absorb. Retrieval-Augmented Generation architecture addresses this gap in part — by keeping the knowledge base current and allowing model context to be updated — but the person framing the right question is still human.What should you actually do? Three steps for Monday morning. First: keep a decision log for your next significant call. Before committing, write down two sentences — ‘Which mode am I in? Am I relying on data, on intuition, or a deliberate combination?’ Those two sentences become the raw material for organisational learning when you revisit the decision six months later. Second: if your team is using an AI-assisted decision-support tool, schedule a review at least every six weeks focused not on what the model measures but on what it excludes. At the SME scale typical of Turkish mid-market firms, this is a feasible exercise with external accounting or advisory support — it does not require a dedicated data science function. Third: when you review outcomes, replace the binary ‘right or wrong’ framing with the question ‘which assumption held and which did not?’ That shift is the actual mechanism through which both the data model and the manager’s intuitive judgment improve over time.The Kayseri general manager eventually brought a revised version of that project to market two quarters later — after observing the competitor’s move and adjusting the concept. His instinct had stopped him; but because he had not recorded why, he could not fully trace what he had learned. The right balance between data and intuition is not a talent. It is a practice. And like any practice, it only converts into institutional memory when it is deliberate, traceable, and repeatable. Knowing which mode you are in when you decide matters at least as much as the decision turning out to be correct.

This article was originally published in Turkish by Gökhan MERCANOĞLU on January 23, 2024. The English edition has been reviewed and edited by the author.


data warehouse creates lasting value only when user behavior, executive ownership, and data quality are handled together. Technology does not create transformation by itself; it only makes the need for transformation more visible. Success is less about the system working and more about the organization learning to work with it.


Gökhan Mercanoğlu
İş Zekâsı ve Raporlama