Weak signalsWeak signals are crucial for identifying early symptoms of strategic discontinuities, providing early warnings of potential threats or opportunities. However, their identification is limited by cognitive biases, lack of data-driven insights, extensively rely on manual work and the absence of systematic methodologies, necessitating advanced techniques and collective approaches for effective detection from multiple perspectives. This study addresses these challenges by utilizing advanced natural language processingNatural language processing techniques and large language modelsLarge language models to create a quantitative driven methodology identifying weak signalsWeak signals from textual data. To operationalize the methodology, weak signals are explored within the context of water networks by examining approximately 90,000 scientific publications and leveraging two key indicators of emergence: intensity and diffusion. This study contributes to theory and practice by introducing an agile and scalable large language model-basedLarge language models methodology for identifying weak signalsWeak signals, enhancing quantitative foresightForesight, and enabling firms to better anticipate and respond to emerging trends and discontinuities.