Berdiyar Saurbayev, Managing Principal, Energy Consulting at EPAM, presents an approach to AI-native enterprise twins that starts with the work an organisation needs to perform. He explains why combining a 3D model with asset information is only a foundation: lasting value comes from using that information within operational workflows.
The presentation explores applications ranging from shift handovers and work preparation to production data verification and hydrocarbon allocation. Berdiyar argues that each use case needs data organised with enough engineering meaning for AI to understand equipment, connections and processes. A searchable collection of documents may help people find information, but more complex tasks require a richer understanding of how an asset operates.
To scale these capabilities, he recommends building reusable components around shared needs. A scheduling or optimisation capability, for example, could support maintenance, turnarounds and drilling activities. This approach allows organisations to develop a roadmap of use cases while reducing repeated development. He also highlights the importance of assessing the total cost of ownership, as a working prototype still requires further engineering before it is ready for production.
As AI agents become part of these workflows, Berdiyar discusses the controls needed to manage them. Organisations must be able to observe their behaviour, define what they can access and do, monitor performance and establish where human intervention is required. He advocates open architectures that allow tools and models to change while organisations retain control of their data, context and workflows.
Watch the full presentation to explore Berdiyar’s perspective on building and buying digital capabilities, and the foundations needed to develop AI-native enterprise twins at scale.
Speaker
- Berdiyar Saurbayev, Managing Principal, Energy Consulting, EPAM
