Takeaways from EAGE Digital 2025
The EAGE digital conference is an annual focal point to discuss key topics including artificial intelligence, digitalization technology and strategies, OSDU, and more.
The fifth edition of the conference was held at the end of March and gathered over 450 attendees of business leaders, technology enthusiasts and experts in energy. With the stunning city of Edinburgh as a backdrop, the conference was friendly and social, providing a fantastic opportunity to connect with many of our clients and industry partners.
The main theme of the conference centered around a crucial question: how can digital technologies empower us to predict more reliably and invest more wisely? As we navigate a landscape increasingly defined by advancements in artificial intelligence, data accessibility, and interoperability, these discussions are more important than ever.
I felt incredibly fortunate to present in not just one, but two sessions, sharing my insights and engaging with fellow attendees. In the following sections, I’ll dive deeper into the key themes explored at the conference and reflect on the insights from these conversations.
Geospatially aware AI for identifying underperforming wells
The first presentation I was a part of showcased an exciting collaboration between Cegal, ESRI and AWS, where we developed a geospatially aware AI-chatbot. We concentrated on developing a tool that would help to identify underperforming wells, which is vital for a company to meet economic and strategic goals and to decide how to appropriately respond to dynamic reservoir conditions. However, the tool could be easily used for any number of other workflows.
Leveraging our own Cegal Prizm API, we extracted key data from Petrel and integrated it with AWS cloud services. The data was then made accessible to AWS's Q model, with query results displayed using ESRI’s web-based map services.
The chatbot was able to, through several user prompts, uncover underperforming wells based on their production history as well as identifying potential reasons behind their low performance. For instance, the model suggested a water breakthrough for one of the wells. We were able to verify this finding through the well reports, which had not been included in our initial data set. This is an impressive result of the solution’s ability to not only retrieve relevant data but to also provide useful interpretations for the user to investigate further.
Read more about data science and Cegal Prizm >
Datadialog – chatting with SQL databases
The second presentation showcased the results of a Cegal SEED fund project from last year where we built an AI agent that could query SQL databases. SQL databases are a very common back-end data structure for applications and services covering anything from financial data, domain data or as well as many of Cegal’s products and solutions. We wanted to make a more dynamic way of interacting with data, so that users could ask anything they like through text.
AI agents are a powerful framework for large language models that enable interactions with outside data sources using “tools”. A tool can be anything from using a google search API to being able to run SQL queries against a database. This means that the model is not only able to understand the user query and write an appropriate SQL query but also execute the SQL code against the database and use the returned results.
We were able to utilize these capabilities to build an agent that could chat with data from a demo Blueback Project Tracker database. Tracker is one of our most popular products and is used to control and monitor Petrel data. It includes both project and domain object metadata where users may want to find out which projects were modified when, what Petrel version was used, or which projects contain specific seismic data. Whilst these questions are possible through the existing Tracker web services, we wanted to test the ability of the agent to navigate the database successfully.
Initially, we found that the model was effective at asking simple questions that only involved one table of data yet struggled with more complex queries. Through adding more information to the prompt, and giving the agent access to more custom tools, we were able to significantly improve its ability to answer more complex questions.
Getting value out of digitalization projects
Many of the discussions at the conference centered around how to effectively productionize digitalization projects and how to make use of new technologies. If we, as an industry can effectively capitalize on new technologies, then we can make better informed decisions and reduce risk.
Enabling innovation and accepting risk
Whilst only around 13% or POCs make it to production, it is still important to support innovation. We can learn from and accept failure but also learn how to be better at shining a light on our successes. This could be helped by flipping the ‘fail fast’ mentality to a ‘succeed fast’ mentality. The energy industry is accustomed to risk management and investment of large sums of money; however, digitalization projects are often treated with a greater perception of risk.
The benefits of AI can enable faster prototyping and faster development times. From a panel discussion on AI, it was noted that one of the biggest barriers to success was the immaturity of the data estate and that this needs to be addressed before we can start to gain real value from AI. A phrase that really stuck out was ‘AI for data and data for AI’, which highlights the need for good quality data, but also the potential for AI to help improve the management of data.
Collaboration and partnerships are key to success
A clear message from several talks was the benefits of collaboration and partnerships between companies, but also within companies.
For example, ‘Fossil vision’ was a project from BP using machine learning image classification to partially automate biostratigraphy workflows for the identification of marker species. The presenters showed the project from the perspective of both the subject matter expert and the data scientist. This highlighted the collaborative effort and need for good communication between experts from different fields. They also showed how to appropriately translate the accuracies and uncertainties of the model into the user interface so that the SME could more easily and efficiently process the data.
The most successful solutions using new technologies such as generative AI or OSDU involved collaborations between vendors, operators, and other partners. This was also highlighted by our own presentation with AWS and ESRI.
Lastly, there is a lot that we can learn from other industries such as finance, power & utilities or healthcare, that are to varying degrees further along the digitalization path and have been faster at adopting AI.
Summary
In conclusion, the EAGE digital conference is a great platform for sharing groundbreaking ideas. The central themes for this year highlighted the critical components for successful adoption of digital technologies: a willingness to experiment and innovate with new tools or ideas as well as an openness to collaboration and partnerships. One of the persistent bottlenecks is the immaturity of the data estate in terms of data quality and data accessibility. There are clear roles for both AI and OSDU in the future and this will take time before the higher impact use cases make it to production.
There are lots of positives for Cegal. We are well placed to work with partners and collaborate with our clients, as we already do. Our presentation with AWS and ESRI highlights how we can move quickly and effectively with other vendors. We also have huge support for our clients in data management through our products, consultants and cloud solutions.
