Energy leaders expect AI to pay back—but the survey did not measure returns
An ADNOC, XRG and Microsoft survey of almost 400 senior energy and finance leaders found agentic-AI deployment rising and grid confidence split. Expectations and reported deployment are not audited savings, reliability or emissions outcomes.
The short version
- What happened
- The company-sponsored survey covered almost 400 senior energy and finance leaders; the accessible release does not disclose an exact denominator, field dates, weighting or response rate.
- Who it affects
- Workers may see more robotics and agentic systems in hazardous operations, while households and businesses depend on grids absorbing new demand without higher costs or lower reliability. Neither outcome was measured here.
- What remains unproven
- Company-sponsored survey
Keep in mind
Company-sponsored survey
Almost 400 senior leaders
Exact denominator not disclosed
No public response rate or weighting
Self-reported deployment
Expected rather than audited returns
No measured reliability, safety or emissions outcome
Source analysis · 1 linked source. Primary analysed source date: 29 October 2026.
Read the full report · 4 min The evidence, methods, implications and limitations
The survey says deployment moved beyond isolated pilots
Energy and finance leaders say AI is moving deeper into operations, but the newest headline figures remain self-reported expectations. ADNOC, XRG and Microsoft surveyed almost 400 senior leaders for the third annual Powering Possible report. Thirty-three per cent said their organisations deployed integrated agentic-AI workflows, up from 15% in 2025, while the share still piloting isolated applications fell from 35% to 15%.
Among energy-sector respondents, 60% said they were piloting or deploying autonomous and robotic systems in physical operations, including infrastructure inspection and hazardous tasks. Sixty-two per cent ranked AI among their two leading corporate priorities, up from 42%, and 89% reported higher investment over the previous year. These answers show executive direction and claimed deployment; they do not reveal how many systems reached full production, how often people intervened or what happened to safety and workload.[1]
Expected payback is not the same as audited return
Sixty-three per cent expected returns on existing AI investments to cover or exceed their costs, rising to 89% for future investments. The wording matters: the accessible release reports expectations rather than audited return on investment. It supplies no project-level costs, realised savings, counterfactual comparison or time horizon. Supply stability and reliability replaced cost saving as the most frequently chosen leading contribution, rising from 26% in 2025 to 39%, but no outage, maintenance or reliability measure is published.
The sponsors have material interests. ADNOC is a major energy producer, XRG invests across energy and technology, and Microsoft sells AI and cloud infrastructure. Their involvement does not make the responses false, but it increases the importance of a full methodology and independent outcome data. The public release says 'almost 400' leaders rather than an exact count and does not expose field dates, sampling, response rate, weighting, questionnaire wording or uncertainty intervals.[1]
The distinction that matters
Three layers of the energy-AI claim
Leadership surveys can show priorities and reported adoption, but outcomes need operational records.
Reported
Leaders describe deployment and spending
Adoption signal
Expected
Respondents anticipate payback and rising demand
Forecast and confidence
Needed
Audit costs, energy, reliability, safety and emissions
Real-world outcome evidence
Grid confidence shows where the physical constraint sits
Seventy-four per cent expected AI to increase energy use by 2030, while only 52% believed existing infrastructure could meet the added electricity demand. Confidence varied by market: 74% in Brazil and 70% in India, compared with 40% in the United States and 38% in the United Kingdom. Respondents most often prioritised grid modernisation and smart-grid technology at 67%, followed by storage at 42%. Those are preferences, not funded capacity additions or forecasts of actual demand.
For communities near data centres and energy infrastructure, the practical questions are whether new loads raise bills, delay connections, increase emissions or improve reliability and local employment. Confidence should change when sponsors publish the full country denominator and methods, and when operators disclose metered energy, outages, maintenance, emissions, injuries and realised financial returns before and after deployment. Until then, the survey is a useful map of leadership expectations—not evidence that agentic AI has delivered safer, cheaper or cleaner energy.[1]