Recent advancements in foundation models like GPT- 4o, Florence, SAM2 are revolutionising computer vision. Traditional computer vision tasks often required custom training for specific applications, consuming significant time and computational resources. Foundation models, however, are large-scale, pre-trained models that serve as a universal base for a wide range of tasks. These models can be fine-tuned with minimal data, enabling faster deployment. The talk will highlight how foundation models streamline the automation of complex workflows, reducing manual intervention and enabling scalable solutions. We discuss real-world applications and understand the future potential of combining vision-based models with broader LLM systems to automate and enhance visual data processing at scale.