JACOPS NV is looking for an AZURE CLOUD ARCHITECT. A classic scenario: a reputable company, a complex task, requiring an experienced specialist to bring order. The job description reads as if they are seeking a hero who will single-handedly bear the digital firmament on their shoulders. Let’s delve into whether a human is truly needed here.
The Belgian multi-technical giant JACOPS NV is a serious player in the world of infrastructure, energy, and telecom. This is not a startup that opened yesterday in a garage. They have complex, long-term projects. This means their IT infrastructure is not a sandbox, but rather a complex industrial mechanism where the cost of error is high.
Judging by the description, the company has a classic set of “pains” of a growing business in the cloud. They have a complex multi-tenant environment. Imagine a digital metropolis with dozens of districts (projects, teams, external partners), each with its own rules. Without centralized management, this quickly turns into chaos: security holes, uncontrolled cost growth, incompatible technologies. Essentially, they are looking for a digital sheriff who will establish laws (governance), design secure districts (landing zones), maintain order (security), and prevent budget waste (FinOps). It’s an important, responsible job. And incredibly routine.
What if this sheriff isn’t a human, but a neural network? What if, instead of one architect who can get tired, go on vacation, or simply miss something, an impartial and tireless AI agent stands guard? Instead of looking for a person to manually “challenge” vendors, a system could be implemented to do this automatically, 24/7, based on data, not intuition.
Let’s imagine a dialogue in a JACOPS meeting room. On one side – the Skeptic, an old-school team lead. On the other – the AI Evangelist, our imaginary manager.
Skeptic: Listen, we need someone with experience. Who will design our Azure landing zones? Who will think through the architecture? A machine isn’t capable of that; it lacks strategic vision.
AI Evangelist: And why does it need vision? We have best practices, such as the Azure Well-Architected Framework. We can take a large language model, trained on thousands of successful architectures and all Microsoft documentation. It will analyze our requirements – security, scalability, cost – and generate several architecture options in the form of ready-to-use Terraform or Bicep code. The human will only need to choose the best one and click “deploy.” This isn’t a replacement for strategy, but its instant embodiment in code.
Skeptic: Okay, let’s say so. But governance! All these policies, tags, naming conventions… Who will ensure that hundreds of developers, ours and partners’, follow all of this? The architect attends meetings, explains, checks…
AI Evangelist: And AI will do it in real-time. We implement AI-driven governance. Instead of writing rules in Confluence, we describe them in code using Azure Policy. An AI assistant in the CI/CD pipeline won’t miss a single commit that violates a naming standard or lacks the necessary tags. Defender for Cloud, with its AI algorithms, will scan the environment 24/7 for policy compliance and automatically correct minor violations. A human architect might make an exception for a friend; AI never will.
Skeptic: Security? Zero Trust, PIM, Conditional Access… this is fine-tuning that requires understanding context.
AI Evangelist: Exactly! And who better to understand context than AI, which analyzes millions of signals per second? Instead of static rules like “this department can, this one cannot,” we get dynamic security. Microsoft Sentinel and Entra ID use machine learning to analyze user behavior. If a developer from Belgium suddenly tries to access data from an atypical location at 3 AM, the system will instantly raise the risk level and request multi-factor authentication or temporarily block access. It will also suggest optimal JIT (just-in-time) permissions, analyzing which accesses an employee genuinely needs to perform a task and which are excessive.
Skeptic: Alright, I’m almost convinced. But how do we get people to trust this? They’re used to having a live person they can ask questions.
AI Evangelist: Gradually. We’ll start with “advisor mode.” The AI system doesn’t make changes itself; it only creates Jira tickets or Pull Requests with suggestions: “I found an suboptimal resource; replacing it will save 200 euros per month. Here’s the code for the change. Approve/Reject?” When the team sees that 9 out of 10 suggestions are useful, we can transition the system to semi-automatic mode for low-risk operations. Transparency is key. Every AI action is logged and accompanied by a clear explanation of why a particular decision was made.
To ensure our AI architect is working as it should, we won’t ask it how it feels. We’ll look at the numbers.
1. Compliance Score metrics in Microsoft Defender for Cloud. Are they growing?
2. Dynamics of expenses in Azure Cost Management. Are anomalous spikes decreasing? Have costs become more predictable?
3. Time to deploy a new environment for a project. Has it decreased from weeks to hours?
4. Number of security incidents related to misconfiguration. Has it decreased?
5. And finally, we’ll order an external penetration test. Let “white hat hackers” try to break into our AI-managed environment. That will be the best exam.
JACOPS is looking for a reliable and experienced employee. And that’s understandable. But perhaps the most reliable, experienced, and tireless “employee” for such a role is no longer a human, but a properly configured and trained set of algorithms. And the human will be left with the more creative task of teaching and guiding this AI.
Источник: https://www.linkedin.com/jobs/view/4411883369/
