Trust by Design: Deploying Responsible AI in Upstream O&G Audit Use Cases
Responsible AI becomes most valuable when it is designed around real business risk, not added as an afterthought. In this presentation, Sofia Silvestre and Edgar Leitão of Galp explore how AI can strengthen upstream audit processes while keeping trust, oversight and accountability at the centre of deployment.
The session focuses on Galp’s role as a non-operating partner in upstream assets, where audit rights provide a vital mechanism for assuring compliance, validating costs and protecting value across joint operations. Rather than approaching AI as a broad experiment, the presentation shows how Galp identified upstream audits as a high-impact area where automation, analytics and responsible AI could deliver measurable improvement.
A key part of the presentation is Galp’s digitalisation journey through Cyclops, an AI-powered audit solution designed to improve document handling and audit effectiveness. By using OCR and intelligent document processing, Cyclops transforms large volumes of supporting documentation into structured outputs, helping auditors move away from repetitive manual work and towards higher-value analysis, recommendations and risk-focused insights.
The presentation then looks at the next stage: scaling from OCR to Gen AI reasoning. This shift brings new opportunities, but also makes responsible AI essential. Galp outlines the practical controls needed to build trust, including evaluation before deployment, safety guardrails, AI reviewers, versioned model runs, prompt injection protection and human-in-the-loop oversight.
The message is clear: AI can accelerate audit work, but the human remains accountable for the conclusion. The full presentation offers a practical view of how Galp is deploying responsible AI in upstream audit use cases, with trust designed into the process from the start.
