As drilling operations become increasingly automated and AI-driven, maintaining effective human oversight is more important than ever. Modern drilling automation systems can support or even perform complex operational decisions, but safety, trust, and accountability still depend on humans understanding what the system is doing and why.
The work examines the key risks, requirements, and constraints associated with enabling human oversight in complex automation systems. It focuses particularly on the growing use of Artificial Intelligence (AI) in drilling automation, where operations are characterized by uncertainty, limited observability, and unpredictable events. Because drilling activities involve high-risk situations that can result in significant hazards and damage, it is essential that human operators receive sufficient and relevant information to understand, verify, and intervene in automated decision-making when necessary.
A major challenge arises from the distributed and multi-vendor nature of drilling automation systems. Information is often decentralized across multiple agents and platforms, making human oversight more difficult. While interoperability may allow direct access to some information, other critical insights may need to be inferred or estimated, adding complexity and uncertainty.
At the same time, system designers must avoid overwhelming users with excessive information. Information delivery should therefore be tailored to different user roles. Operational users may need information about data quality, uncertainty, and model estimates, while supervisory or monitoring personnel may require insight into interactions, dependencies, and coordination between automation agents. Ongoing work with an explainability screen is going to be tested by drillers in OpenLab with personnel involved in the planned offshore test on DeepSea Stavanger.
Key takeaway: As drilling systems become more autonomous and AI-driven, successful human oversight depends on providing the right information to the right users at the right time. This requires balancing transparency, explainability, uncertainty awareness, and information management in complex, distributed automation environments.
Collaboration with HAVTIL
We are very pleased with the collaboration with HAVTIL on this work. To have more information take a look at:
- HAVTIL’s websites When AI makes the decisions, and
- The article A Technical Perspective to Human Oversight in Complex Drilling Automation Systems Based on Artificial Intelligence Methods authored by R. Mihai, NORCE; L. I.V. Bergh, HAVTIL; E. Cayeux, NORCE; B. Daireaux, NORCE; E. Lootz, HAVTIL