It's also important to understand that ML cannot be used in all situations to support complex operational decision making. To incorporate ML, it's important to have good quality data. However it has been observed that, many organisations lack data maturity and hence to use ML, it's important to digitise the processes first.

El uso de técnicas de machine learning las uso para hacer predicciones de variables con tendencia y estacionalidad, para esto es muy importante la recolección de datos históricos. Los modelos de ML permiten tener resultados más precisos y ahorro en el tiempo de análisis, permitiendo hacer foco en la calidad de los datos con el que se alimenta el modelo y en los resultados para responder las preguntas del negocio basada en datos. Esto se conoce como “Data Driven”

ML can help derive new insights which is then fed into the operational engine as new data points to optimise decision making. For example, in a resource allocation problem, ML algorithms can be used to predict the performance output of the targets, then the allocation of resources can be adjusted to optimise the overall output. In a scheduling problem, ML methods can be employed to predict availability of resources, job completion time or future demands. The estimation is then used as inputs to formulate and solve the scheduling optimisation problem.

My favourite tool is the effort-benefit matrix, which we've extensively used at Generali Česká pojišťovna. It helps us prioritise ideas at every stage: from initial scoping to proof of concept (POC), and then again for features within those POCs. It's beneficial for teams lacking advanced skills, fostering valuable discussions and aligning everyone involved. Initially, estimates are based on intuitive "t-shirt sizing," later, they become more precise, with data-backed benefits for specific features.