Sunya | Energy AI Observatory · Open the interactive dossier
9 AI initiative profiles for Chevron across 7 departments, with evidence grades and source links. Strongest evidence: B · Primary publication — ApEX multi-agent exploration search.
Corpus size: >1 million files / 4 TB (Company-reported)
No quantified results disclosed
5% lower LOE; BOE output about 6% above original projection after normalizing for new-drill wells, in first year.
No quantified results disclosed
Not separately disclosed
Over half of employees have tool access
Potential billions annually
~2.67 GW; 20-year PPA; first power targeted 2028
Supplementary notes can update existing programs or describe adjacent work. They are not extra adoption counts.
Chevron says an internally built production assistant works continuously in the Permian, Bakken and Partitioned Zone.
No project-specific benefit disclosed.
Limits: Automatic equipment-write authority and override rules are unstated. The aggregate AI value claim is not allocated to these tools; no net ROI established.
Chevron Q3 2026 earnings call: Chevron-edited transcript · Unknown
Chevron says AI helped anticipate downtime in the DJ Basin, applying lessons from the Permian.
The figure counts events, not production volume or downtime hours. Absolute counts, exposure and the comparison period are not given.
Limits: No costs, attributable cash benefit or net ROI. The separate up-to-23% chemical-treatment production claim is not this AI result. Do not equate fewer events with a 50% production gain.
Chevron 3Q2026 earnings presentation with prepared remarks · 2026-10-30
SPE reports that Chevron’s Peter Nesom described GEORGE, a production-engineering assistant used for analysis, optimization and automation tasks on more than 5,000 wells.
More than 5,000 wells is reported use scope. No production uplift, time saving, cost or net return is supplied.
Limits: Original conference reporting of named operator remarks, not an operator-authored deployment paper. The event program corroborates speaker and session only. Keep as reported use awaiting technical corroboration, outside additional adoption/result counts. Do not infer an LLM architecture, autonomous field control or a company-wide knowledge system.
Maximizing Every Barrel: Artificial Lift Embraces Automation · 2026-09-04
SPE Artificial Lift 2026 opening plenary program · 2026-08-25 (event date; page publication date unknown)
Full transcript screen · call 2026-10-30
Complete 17-page Chevron-edited original transcript topically screened, including prepared remarks and Q&A through call close. Relevant AI, software, automation and power-demand passages read in context. Text reader succeeded through the official event-page link; direct fetch and rendering failed. No replay listening claimed.
Prepared remarks describe DJ downtime prediction, ApEx data retrieval and ScoutAI assessments across ten basins.
Reported AI value exceeds $100m year-to-date; over 150 exploration leads yielded 16 prospects for evaluation.
Limits: Aggregate gross/net basis and costs unspecified. Leads are not discoveries. DJ event reduction already logged separately.
In-house production assistant deployed in Permian, Bakken and Partitioned Zone; well-design and capital-scoping tools remain in early tests.
No project-specific outcome disclosed.
Limits: No well count, control permissions, comparison or costs.
Human-agent teams support turnaround and refinery planning; AI also informs reliability and yield analysis.
No isolated benefit disclosed.
Limits: Catalyst-driven yield gains are separate from AI.
Kilby has a Microsoft power agreement; permits, financing and final investment decision remain pending.
2.67 GW; 20-year PPA.
Limits: Contracted power is not operating AI adoption or cash return.
Chevron Q3 2026 earnings call: Chevron-edited transcript · Prepared remarks PDF/printed pp. 2–5; Q&A pp. 6–17 through call close. New workflow-stage distinctions p. 10.