Over the past few years, Statoil has explored how multi-agent designs, platforms, and autonomous-system capabilities can be used to improve core upstream oil and gas business processes. Our hypothesis is that autonomous-system properties such as collaboration and negotiation can be used to improve human decision making and thereby enhance both operational efficiency and safety. To most of us, an autonomous system in line with Stanley Kubrick’s HAL 9000 from 2001: A Space Odyssey is unacceptable. Therefore, an overarching principle in an architecture is that the human operator must be in control and that the system should be regarded as a useful servant. To support this, we have developed an architecture whose bearing concepts are variable and delegated autonomy. In this presentation, we will discuss the principles, experiences, and challenges encountered with this architecture.