Quick answer: Data, and a lot of it. Sergei Ivanov opened the ENERGYai announcements expecting to read about agents and mostly read about data: petabytes that ADNOC collected over decades before the first agent was trained. That’s good news, because data is something an operator can start on today without waiting for the perfect model. Above the ground, that data lives in drawings. At Future AI Middle East in Abu Dhabi on 29 October, DIGATEX and Azati showed what happens once drawings become data: engineers ask the DIGATEX P&ID agent a question in plain language and see the answer highlighted on the source sheet. One thread ran through almost every conversation that day. Everyone wants agents, and the real work starts underneath them. Which drawing has given your team the most trouble?
The shallow lesson: wait for a better model
The model is the part you can buy next year. Your data is the part only you can prepare.
The real lesson: the agent is the last step, and the data work can start now
ADNOC’s agents sit on decades of collected data. For plants, the equivalent is the drawing set. Turn it into structured data and an agent has something to reason over. Skip that step and it’s guessing at pictures.
The sobering part
Operators in the room didn’t need more AI demos. They needed AI they can trust on their own data, and trust comes from seeing where each answer came from on the sheet.
Full disclosure: Azati has been DIGATEX’s AI R&D lab since December 2017, and we were a Silver Partner of the conference together with DIGATEX. We are not neutral on “data first.” See our engineering AI work.
FAQ
Adapted from LinkedIn posts by Sergei Ivanov, CEO, Custom Software & AI Solutions in GCC – read the original – and by Azati Software, written around Future AI Middle East 2026 (Cavendish Group), Abu Dhabi, 29 October 2026.
