Quick answer: Eleven. Sergei Ivanov took Azati’s P&ID agent to GCC AI Digital in Oil & Gas 2026 with a narrow plan: read P&IDs and drawings, extract data, turn static technical documents into structured data. At the booth, people handed him eleven problems: fragmented engineering data, complex technical documents, extraction, P&IDs and drawings, asset context, well information, inventory and MRO, digital twin readiness, their own past attempts to build something similar, information search, and turning documents into structured data. So he did the nerdy thing: split the list into two piles – what Azati can already prove, and what is genuinely new. Which pile is your problem in?
The shallow lesson: one agent can do everything
A long wish list isn’t a product spec.
The real lesson: eleven problems, one root
Almost every item traces back to engineering knowledge locked in documents. Structured data is the precondition for digital twins, MRO, asset context, and search. Solve the extraction well, and the rest stops being eleven separate projects.
The sobering part
Many of the people at that booth had already tried to build something similar themselves. The hard part was never the demo; it’s reliability across messy legacy documents at real scale. Sorting “proven” from “new” honestly is the only credible way to answer.
Full disclosure: it’s Azati’s agent, the same technology that processed 120,000 engineering documents in under 24 hours at 98.8% accuracy, so we’re biased. See our engineering AI work.
FAQ
Adapted from a LinkedIn post by Sergei Ivanov, CEO, Custom Software & AI Solutions in GCC – read the original, written after GCC AI Digital in Oil & Gas 2026, organized by The Resources Network UAE.
