Complex decision making in energy companies is often slow, iterative and based on incomplete analyses. Because of this, value is left on the table. Root causes for this problem include outdated toolsets, siloed and hierarchical organisational model and old-fashioned linear workflows. This does not only lead to suboptimal decisions, but also to low employee engagement. This presentation will explain how the latest Artificial Intelligence, famous from achieving superhuman performance in games such as Go, Starcraft and Poker, can be used to transform decision-making. A real use case, AI-assisted Well Trajectory Planning, will be presented to show how digital solutions can help to break through the organisational silos, enable better and faster business decisions and make technical work more fun. This use case will stress a number of key enablers for successful digitalisation: translation between domain experts and data scientists, cross-disciplinary integration skills, early adopters and believers  Key challenges that will be highlighted include data access and consistency and asset-specific challenges that lead to a dilemma of custom-built solutions vs fast replication and scalability.

At Neptune Energy we are well underway with building a digital culture and transforming our company. Our Digital Subsurface Initiative is driven by the business in tight collaboration with IT. We have focus on bringing R&D into the workflows fast and we do a lot of testing to make the right decisions. We have internal and external Dev Ops teams including data scientists, data engineers and software developers to work on our use cases, projects and related data management challenges. One of our use cases that we have developed together with our IT provider Cegal is called Evergreen Maps. Seismic interpretation sitting in the different users’ work projects are made available in one GIS view. In addition Evergreen Maps enables easy production of regional maps using the best interpretation available at all times. This saves time when starting new projects in new areas, screening for new opportunities and keeping our data tidy. The use cases in our portfolio range between pragmatic digitalisation projects to high potential machine learning R&D projects. We have high ambitions and have stated that we want to get from idea to discovery 70% faster. The first reaction to that from the subsurface staff was lukewarm, then we experienced that people just needed some time to think and they started to come up with ideas on how to make it happen.Currently, we are at the phase of deployment for our first use cases and we are working on the different challenges that are surfacing around data flow and compute. In addition we have started to introduce agile ways of working to our teams, because new tools only will not take us to the level of our ambitions.

It is difficult times for the oil and gas industry, which is why companies are implementing machine learning technologies to drive efficiency throughout the organization. But success has been limited and one of the key reasons, which are often overlooked, is the inability to access clean, structured high-quality data. The other factor that affects the success of such initiatives is the ability to apply the appropriate machine learning techniques within an established geoscience workflow in the organization. Should workflow remain the same in the face of such new technologies, or should they change completely? Looking at the well-known play-based exploration workflow as an example, Iraya’s focus is finding the “path of least resistance” from basin-to-prospect- to- target, by solving the problem at the source- the consolidation and accessibility of prior-knowledge from unstructured data gathered over decades. We then use these as input to petrophysics, geophysics and engineering, which by itself, also requires separate machine learning techniques. The combination of both, unstructured and structured techniques and expert human-in-the-loop for quality control makes for effective end-to-end geoscience workflow enabled by ML.