The Paranal Observatory is the host of VLT/VLTI, a world-class telescope that contains systems with different maturity levels. In the middle term Paranal will operate ELT and CTA. We are exploring new techniques to deal with this increasing complexity. A good part of system behavior is well captured by software logs, and several techniques brought from academic literature and as industry practices have been implemented as tools for system monitoring, error profiling, anomaly detection, and process discovery among others. Also, several prototypes have been made using ML and natural language processing techniques with the aid of Paranal DATALAB, a Docker-based data warehouse that provides years of historical data. This work shows the result of our experience and identifies key challenges for the adoption of log analysis techniques in the operation of other astronomical facilities.