Quick answer: Not yet, for most operators. ADIPEC 2026 ran in Abu Dhabi on 2–5 November, and its AI Zone, co-hosted by AIQ, showed AI on real tasks: reservoir modeling, autonomous control, infrastructure inspection, subsea robotics. Digital twins were everywhere, from vessel simulators to full-scale models of energy plants. But in almost every conversation Sergei Ivanov had on the floor, one question came back: how do we get decades of engineering data into a form AI can actually use? A digital twin is only as good as the drawings and plant documents underneath it. So before you ask what the twin can do, ask what it will be built from.
The shallow lesson: AI in energy is finally real, so go buy some
The demos are real. They also run on data somebody spent years structuring first.
The real lesson: the twin is the visible part; the data is the work
Reservoir models and autonomous control impress on a stand. What decides whether they work on your asset is whether your P&IDs, isometrics and plant documents exist as structured, connected data or as scans in a folder.
The sobering part
Most plants were designed long before anyone planned to feed them to a model. Decades of revisions, mark-ups and contractor formats don’t convert themselves, and no agent can reason over a drawing it can’t read.
Full disclosure: Azati turns engineering drawings and plant documents into structured data, so we would say the foundation matters more than the demo. See our engineering data work.
FAQ
Adapted from a LinkedIn post by Sergei Ivanov, CEO, Custom Software & AI Solutions in GCC – read the original, written on day 3 of ADIPEC 2026 in Abu Dhabi. Source: WAM, “ADIPEC 2026 maps path from AI ambition to industrial-scale deployment”, 5 November 2026.
