AES has been running safety audits the hard way for a long time. As a global energy company, it runs more than 1,500 safety management program audits per year. Each one consumed roughly 100 employee hours: people manually reading hundreds of pages, cross-referencing standards, generating compliance protocols. That’s not a workflow. It’s a tax on everyone involved.
The question wasn’t whether to change it. It was whether AI could handle something this consequential, at this volume, with the accuracy a safety-critical operation actually requires.
Two Months from Concept to Production
AES partnered with Google Cloud and Anthropic to find out. The result is an agentic system running Anthropic’s Claude models on Vertex AI. Claude’s long-context capability handles up to 400 pages at a time, and its multilingual support made it the right fit for a company operating across North and South America. Vertex AI Model Garden simplified the integration. Because AES already had a relationship with Google Cloud, connecting the two didn’t require reinventing the data infrastructure. According to Marwan Sherri, Data Scientist at AES, it took just two months from concept to operational agents.
That timeline matters. Enterprise AI projects in safety-critical environments often stretch for years before reaching production, weighed down by compliance reviews, integration complexity, and institutional caution. AES moved in two months. The speed reflects both the maturity of the tooling and the clarity of the use case.
The Numbers
After running more than 50 agent audits, the results are hard to argue with. Audit costs dropped by 99%. Results that previously took 14 days now come back in one hour. With AI agents handling half the workload, AES can now run double the number of audits annually. Audit accuracy improved by 10 to 20%, which matters enormously when what you’re auditing is safety compliance at energy facilities. More audits, faster, at higher accuracy, for a fraction of the cost: that’s a genuinely different operational reality.
The Anthropic angle is also worth noting. Claude is available through Vertex AI Model Garden alongside Google’s own models. Deployment, security, compliance, and access management all stay unified within a single GCP environment. AES didn’t have to choose between a best-in-class foundation model and enterprise infrastructure. They got both.
Want to go deeper?
- AES customer story (Google Cloud), The full case study covering the architecture, outcomes, and partnership details.
- Vertex AI Model Garden, Where Claude and other frontier models are available with unified enterprise governance on Google Cloud.