Geosteering workflows are increasingly based on updated quantifications of subsurface uncertainties during real-time operations. These workflows give tremendous amounts of information that a human brain cannot make sense of. To advance value creation from geosteering, the industry should develop and adopt decision support systems (DSSs). DSSs might provide either expert tools which inform decisions under uncertainty or optimization-based recommendations. In both cases the adoption of a DSS would require new skillsets to dynamically and systematically interpret uncertainties and parameters required for operational decision making. The aim of this work is to identify the relevant skills and ways to aid good geosteering decisions.
We present an experiment where 54 geosteering experts took part in performing steering decisions under uncertainty in a controlled environment using an online competition platform. In the experiment we compare the decisions of the experts with an AI bot that had the same information at its disposal.
Two of the participants beat the AI bot. A survey was conducted to reveal their winning strategies. The survey shows that both of the winners had extensive prior geosteering experience. That, together with luck, allowed them to beat the AI bot. At the same time neither of the winners utilized the full potential of uncertainty tools in the platform.
While geosteering experts possess insights due to prior experience, the information in the real-time data will still be overwhelming, sometimes resulting in inconsistent and unreliable geosteering choices. The AI bot guarantees reliable and consistent decisions by optimization based on systematic uncertainty analysis. Further development of DSSs, and their use as training-simulators for experts, should lead to improved well placements through adopting well-established principles for high-quality decision-making.