In January 2011, IBM’s system called Watson defeated Jeopardy champions Ken Jennings and Brad Rutter. The broadcast is still weeks away, but the result is already circulating in technology circles. Many executives are filing this away as a curiosity — a clever demonstration of computing power. That reaction misses the point entirely. Watson did not run a database query. It parsed natural-language clues, weighed ambiguity, and ranked candidate answers by confidence score. That distinction marks a meaningful threshold for anyone running enterprise decision support systems.
Cognitive computing differs from rule-based systems in a fundamental way. Conventional decision support software operates on structured data and predefined logic: if sales fall below a threshold, trigger an alert; if inventory drops to a set level, generate a purchase order. Watson works differently. It processes unstructured text — human language — identifies contextual relationships across vast document sets, and assigns probability weights to its outputs. This is categorically different from what an ERP reporting module does, and the gap matters for how we think about the next generation of enterprise software.
For mid-sized businesses in markets like Turkey, Watson is not a product on the shelf today. It is a direction indicator. The relevant question for decision-makers is not ‘can we buy Watson?’ but rather ‘are we building the kind of data infrastructure that will allow us to benefit from cognitive capabilities as they become commercially available?’ Most honest answers will reveal significant gaps. Supplier correspondence, customer complaint logs, contract documents, quality reports — this unstructured information sits in email archives and paper files, invisible to the structured systems that run daily operations.
The practical business lesson from Watson is about where knowledge actually lives. An ERP system at a manufacturing company keeps production orders, inventory movements, and invoice records in good order. But a decade of supplier communications, warranty claims, and field service notes typically exists outside that system entirely. Classical software cannot see this data. Cognitive approaches are designed precisely to make this hidden knowledge mass processable and searchable. A logistics company trying to anticipate delivery bottlenecks, a distributor assessing supplier reliability, a professional services firm reviewing contract risk — all of these are decision domains where unstructured data processing capacity would create measurable value.
Realism is warranted here. Watson today is accessible through large institutional partnerships and research collaborations, not through a standard software procurement process. The total cost of ownership for deploying cognitive infrastructure at enterprise scale remains well beyond the reach of most mid-market companies. But the underlying technologies — probabilistic ranking, contextual inference, text analysis — are moving toward commercial software products. SAP and Oracle have both signaled investment in this direction. The infrastructure choices made today should therefore be evaluated partly on their capacity to accommodate these capabilities as they mature.
The practical challenge is twofold. First, language: Watson was trained on English-language sources and optimized for English. Building equivalent capacity in Turkish requires separate linguistic work and training data at scale — this is not a trivial adaptation. Second, organizational readiness: cognitive systems produce probabilistic outputs, not binary answers. When a system says ‘seventy-two percent probability of supplier delivery delay next quarter,’ the organization needs a decision process that can act on that signal. If existing workflows are built around deterministic system outputs, the most sophisticated cognitive layer will still underperform because the surrounding process cannot absorb what it produces. This is an organizational design problem as much as a technology problem.
The actionable recommendation for managers is straightforward. Do not archive the Watson story as science fiction. Instead, map the unstructured information in your organization: which business decisions are currently made on intuition because the relevant data exists but is not accessible to your systems? Supplier selection, customer segmentation, contract risk assessment, warranty pattern analysis — identify the two or three domains where unstructured data access would most change the quality of decisions. When evaluating your next enterprise software investment, ask vendors directly how their platform handles unstructured data and what roadmap they have for probabilistic decision support. The companies that ask these questions now will be better positioned when cognitive capabilities become a standard line item in the enterprise software catalogue.
This article was originally written in Turkish by Gökhan MERCANOĞLU on January 17, 2011 and has been automatically translated into English and other languages using machine translation.