AAPG Subsurface AI Special Report: Summary & Review

     This special report was sponsored by Aspen Tech. AI in oil & gas is not new. Beginning with simple neural networks and in the past decade with large language models and generative AI, the sector is embracing AI and achieving positive results. The report cites seismic surveying, reservoir evaluation, data validation, production strategies and optimization, multidisciplinary collaboration, predictive maintenance, and drilling optimization as examples. Digital twins, or real-time digital models, have been and are being deployed in several areas. Digital twins and generative AI have already been credited with increasing production and decreasing costs. Oil & gas companies generate vast quantities of data, including 3D seismic data, well logs, drilling data, and production analysis data. Thus, they are quite amenable to AI analysis that can find hidden relationships in the data.

     The report identifies four key themes for subsurface AI in 2026.

1. Simplifying Vendor Portfolios. This is important for reducing the siloing of data and putting it all in a single system. With a single encompassing platform, AI can act as an application programming interface (API). AspenTech’s Subsurface Intelligence “brings together domain-specific agents (for geophysics, formation evaluation, petrophysics, geomodelling and reservoir engineering) using AI capabilities.”

“AI-propelled software is enabling companies to use a single platform to work with multimodal data. Data analytics software company Databricks’ Data Intelligence Platform bridges data warehouses (where structured data informs decisions, but raw data is often problematic) with data lakes (repositories for raw data). “

“Leveraging Databricks, companies standardize schemas, manage quality, and keep datasets in sync as new sources come online,” says Enterprise Solutions Specialist at DataBricks, Reagan Kennedy. “With that data they can apply AI/ML capabilities, build out analytics, and expose the data via APIs and applications so existing tools can read/write against the same data instead of maintaining their own silos.”

2. Automation of Upstream Workflows with Agentic AI. They define agentic AI systems as systems that “autonomously act, decide, and orchestrate multistep workflows.” These systems are currently moving from the pilot phase to being fully operational. Output moves from making suggestions in generative AI to initiating actions in agentic AI.

“…generative AI might suggest a reservoir model to help with well planning. Agentic AI would use automation to rapidly create the model and then query asset-wide data to understand where the next best places to drill are or to predict the outcomes of different development strategies.”

“Additionally, multi-agent frameworks can coordinate across subsurface disciplines simultaneously: One agent interprets stratigraphy, another models pore pressure, a third cross-references offset well data.”

     Agents are typically limited to what they were designed to do, such as analyze the geology of a single basin. Thus, location-specific and domain-specific validation is important. AI offers faster project analysis and better integration of data, and geologists should incorporate it smartly to save time and improve overall subsurface analysis.

3. The Open Subsurface Data Universe Launches. The Open Subsurface Data Universe (OSDU) was developed beginning in 2018 by major oil companies, including Anadarko, BP, Chevron, ConocoPhillips, Devon, Equinor, Exxon, Shell, and TotalEnergies, along with tech companies. It aimed “to create a single, open, cloud-native data platform where seismic, well log, reservoir, and production data could live, communicate, and be accessed by any application, regardless of vendor or operator.” OSDU was not without rollout problems in integrating different kinds of data, but it is still being perfected. AspenTech’s OSDU czar, Dani Alsaab, noted:

“The future is multi-vendor collaboration in the upstream ecosystem, and we see ironclad commitment to OSDU as a differentiator for the future.” 

     The report notes that OSDU is still the best framework yet devised for the standardization of subsurface data. However, it has yet to be widely adopted outside of large oil & gas companies and still has challenges with interoperability and the integration of proprietary data.

4. Using AI to Train the Next Generation of Workers. They note that new employees can be aided by better AI integration of “how-to” functions in software so that software competency is improved and employee training is advanced. Software competency is a very important skill for modern geologists. However, it should not eclipse training in geology, but ideally complement it.

“GenAI with domain guardrails is the way to transfer technical knowledge and experience to the next generation of workers,” said AspenTech’s AI CTO Heiko Claussen. “We see this as a way to give our four decades of technology leadership that has been built into our domain-specific software a new lease on life: It is an evergreen way to make future workers experts. It empowers us to add value to industry in new ways. There will be plenty of jobs, but the core skills will be how knowledge workers best leverage, interpret, and take advantage of these new AI tools.”     

     While I think this is all important, I also think that geologists should not abandon the basics of geology for AI-based approaches in all domains. We should always remember that AI functions best as a digital assistant, hopefully a remarkably competent one that can advance successful solutions to problems.

    

References:

 

AAPG Subsurface AI Special Report. Sponsored by AspenTech. 2026. AAPG_SubsurfaceAI_SpecialReport_2026.pdf