No black box, but full control over AI
Reliable AI requires control, monitoring, and accountability. Only when there is insight into performance, security, and data flows is AI truly suitable for enterprise environments.

Reliable AI demands control, monitoring, and accountability. Only with deep visibility into performance, security, and data flows is AI genuinely ready for enterprise environments. Azure AI Foundry plays a central role here, offering integrated enterprise controls and observability. For BBTG, this forms the cornerstone of a responsible AI First approach. Production-grade AI solutions must be more than just smart; they need to be predictable, manageable, and traceable. Especially in environments where compliance, privacy, and continuity are essential, visibility is just as critical as innovation.
Enterprise controls: no compromises on security
Every AI solution that BBTG develops and manages meets strict security and compliance requirements. Azure AI Foundry supports this with deeply integrated enterprise controls, including role-based access control (RBAC) at the hub, project, and resource levels, private and managed virtual networks for network isolation, and integrations with Key Vault and Entra ID for identity and secrets.
This allows agents and their data to be precisely defined, fully in line with existing IT and security architectures.
“Security is not optional for us,” says Marco Boele. “It is the foundation upon which sustainable digital innovation can exist, especially in an AI-first combination.”
In practice, this means that AI solutions always:
- Have access only to explicitly assigned resources, following the principle of least privilege
- Operate within defined networks, without uncontrolled data exchange
- Manage accounts and secrets centrally and auditably, in accordance with existing compliance guidelines
This foundation ensures that AI behaves predictably, even when solutions are scaled up or become more complex.
Working transparently with observability
In addition to protection, insight is essential. With the observability functionality of Azure AI Foundry, traces, spans, metrics, and logs are visible via Azure Monitor and Application Insights. This gives developers and administrators insight into the entire lifecycle of AI agents, from input and model selection to actions and output. Every tool call and interaction is traceable. This supports debugging and quality improvement, and makes AI solutions explainable to security, risk, and compliance teams.
Observability also ensures that anomalies are recognized and corrected faster, even before they have an impact on processes or users.
From proof of concept to production, without surprises
Rubicon’s approach is focused on control over the entire AI journey. By combining enterprise controls and observability with mature DevOps and CI/CD pipelines, AI solutions are rolled out and managed in a controlled manner, even in complex and regulated environments. This prevents successful proofs of concept from becoming disconnected from governance and management. AI solutions continue to function predictably, even as usage, data, and responsibilities increase.
Innovation that stands the test of practice
Developments surrounding Azure AI Foundry are moving fast. BBTG continues to build on proven techniques and critically evaluates new possibilities before they are deployed in production.
“Our responsibility is to ensure that customers not only benefit from innovation,” concludes Jeffrey Paarhuis, “but can also truly rely on it. That is the core of ‘AI First, With U and AI’.”
This blog post was previously published on rubicon.nl
Featured posts
We don't do off-the-shelf job openings. We create your ultimate role based on your personal ambitions and expertise.

Snowflake World Tour
What value is your organization getting from data and AI? On October 27, BBTG will be at the Snowflake World Tour in Rotterdam. As a Snowflake Elite Partner, we look forward to discussing what it takes to turn AI into tangible value for your organization.

Using AI isn't that hard, but getting real value out of it is.
Normally, as an IT organization, you have to explain why a new technology or tool is interesting. With generative AI, the opposite happened: employees started using it themselves and immediately saw its value. From creating texts and presentations to analyses and initial drafts of reports. And that creates a new reality.

Sovereign cloud and data sovereignty
This year marked a turning point for the sovereign cloud. More and more governments and companies want—or need—to reduce their dependence on US cloud providers. In this blog, we explore the current state of affairs and the developments you should keep an eye on. We speak with Technical Director Rik van Berendonk, who summarizes how to put data sovereignty into practice.



