Global Screening Services reduces Terraform review time with an AI Cloud Guardian

Storm Reply built an AI assistant for Global Screen Services (GSS) that explained what proposed cloud changes would do and highlight potential disruption, reducing complex release preparation from two days to less than four hours.

The CHALLENGE

Release changes to GSS’s cloud infrastructure faster without increasing the risk of disrupting its sanctions screening platform.

THE SCENARIO

A careful review process that consumed engineering time

GSS provides a sanctions screening platform for regulated financial institutions, where the platform has to stay available and trustworthy at all times. In this environment, every infrastructure change must be assessed carefully.

GSS manages this infrastructure using Terraform. Terraform allows engineers to define servers, networks, permissions and other cloud resources in code. Before a change is released, it produces a detailed plan showing everything that will be added, changed or removed. This required engineers to read the plan and work out whether planned changes could lead to disruption. While a single review would take 30 minutes, a complex release could require two days of work.

The Solution

An AI assistant that identifies disruptive changes

Storm Reply worked with GSS to build Cloud Guardian, an AI assistant integrated into the company’s existing release process.

After a proposed change passed its initial automated tests, the release pipeline produces one or more Terraform plans. Cloud Guardian read these plans together, looked for potentially disruptive changes and creates a concise report for the engineer.

Instead of repeating the technical plan, the report explains the changes, identifies affected infrastructure and highlights any issues that require closer attention.

A recommended route for each change

Routine updates deploy automatically, while high-risk changes are intelligently escalated to a controlled release process for mandatory human review.

Automated checks on the AI's output

Automated validation ensures the AI’s output always aligns with GSS’s release policies. This strict governance prevents teams from acting on incomplete, inaccurate, or non-compliant data.

Earlier visibility of differences between environments

Engineers gain immediate insight into configuration discrepancies between staging and production, allowing them to resolve environment drift before pushing changes live.


HOW WE DID IT

Embedded collaboration and iterative design

Rather than a traditional vendor hand-off, Storm Reply embedded directly within the GSS platform team. This co-development approach ensured the AI model was finely tuned to GSS's specific release pipelines and cost-efficiency requirements.

Because Cloud Guardian was injected directly into existing workflows, GSS didn't need to overhaul its toolchain or introduce friction with new platforms. Furthermore, the embedded approach upskilled GSS engineers, equipping them with the knowledge to maintain and scale the AI solution independently.

THE RESULT

From reading every line to reviewing exceptions

Cloud Guardian changed the shape of the review process. Instead of reading every line of every plan, engineers now focus on the specific changes the assistant flags.

80% faster complex releases

Preparation for complex releases plummeted from two days to under four hours, freeing engineers from the slowest part of the cycle.

90% reduction in routine reviews

Individual plan reviews dropped from 30 minutes to approximately three minutes, as engineers only needed to review the specific risks the AI flagged.

A standard check for every plan

Every Terraform plan now receives a consistent, thorough baseline assessment, neutralizing human fatigue during lengthy release cycles.

A foundation for independent component releases

GSS is working towards releasing safe components independently, so that one problematic component no longer holds up changes that are ready to go.