For organizations adopting AI, getting a pilot to work is only part of the job. Running it in production requires decisions about infrastructure, enterprise data, access controls and how to evaluate performance over time. Those decisions often take more work than the initial prototype.
Today, at the Google Cloud Doha Summit, Scale and Google Cloud are publishing a joint reference architecture for running the Scale GenAI Portfolio (SGP) on Google Cloud, integrated with Gemini Enterprise. It gives technical teams a documented deployment pattern they can use as a starting point, with guidance on how the components fit together.
How the platforms work together
The architecture connects Google Cloud’s infrastructure and services with Scale’s tools for building, evaluating and operating AI agents.
Google Cloud provides access to Google and third-party models, identity and governance services, and infrastructure including Google Kubernetes Engine (GKE), GPUs and TPUs. Gemini Enterprise gives employees a place to discover and use agents within their existing work environment.
Scale provides SGP for agent development, orchestration and deployment, along with evaluation, tracing and human review. Scale’s contribution also includes domain expertise, support for languages including Arabic, and delivery teams that remain involved after deployment.
The reference architecture sets out how these capabilities work together, so teams can plan the full deployment rather than resolve each integration as they go.
From agent development to employee use
An agent is built and evaluated in SGP, deployed within the customer’s Google Cloud project using their VPC and encryption keys. The same agent can then be made available through a dedicated business application and through Gemini Enterprise.
The architecture uses A2A and MCP for interoperability, alongside the Agent Registry for discovery. This approach allows teams to make an existing agent available through another interface without rebuilding its underlying logic.
Supporting Google Cloud services address different parts of the deployment from gateway access to controls for prompt safety and network protection, and operational visibility and records for audit and review.
Together, these components give teams a common foundation for operating agents across applications.
What this means for development teams
Teams can build an agent for a specific workflow and also make it available through Gemini Enterprise. Employees can discover it there, subject to the organization’s access permissions, while the team maintains the same underlying agent.
Standard interfaces also make it easier to connect agents with enterprise data and with one another. Teams can apply identity, data loss prevention, audit and spending controls through a shared Google Cloud governance framework, then address any additional requirements of the individual use case.
This lets customers use Scale’s evaluation capabilities and domain expertise within their Google Cloud environment. It also gives the teams responsible for development, security and operations a shared deployment design to review.
Where to start
Throughout deployments with customers across MENA and globally, we’ve found that success starts with identifying one use case that has a clear owner and a measurable outcome. Then, use the reference architecture to map its data sources, access requirements, evaluation criteria and deployment needs. Establish what acceptable performance looks like before launch and how the team will monitor it afterward.
That work creates a foundation for subsequent deployments. Teams can reuse the infrastructure and controls that fit, while evaluating the requirements of each new agent.
For use cases involving specialized domains, Arabic-language requirements or regulatory constraints, Scale and Google Cloud can work directly with customers, with the published architecture giving those discussions a concrete starting point.
Ready to break through your data bottleneck?
Scale's team will match your project to the right experts, fast.
