Entrazure
Filed by problem, because that is how the call arrives
A product-shaped catalog answers the question only after you already know which product you need, which is usually the part you were hoping to buy an opinion about. So these pages start one step earlier, at the symptom, and the technology decision appears where it belongs: after the problem has been described precisely enough to have a right answer.
It also keeps us honest about scope. Half of these problems turn out to span several services. An unexplained bill is a governance problem before it is a rightsizing exercise. A stalled AI pilot is usually a permissions problem wearing a model’s clothing. Filing them by product would hide exactly the part that makes them hard.
The six
AI agents in production. The pilot demoed well and never shipped. Read this if you can name the use case but nobody can answer who the agent sees documents on behalf of, what it costs per month at real volume, how often it is wrong measured against what, and who owns it after the person who built it changes role.
Microsoft 365 Copilot rollout. The licenses are bought and usage is flat. Copilot answers from whatever a person can already open, and in most tenants that set is wider than the org chart would suggest, so this one is a permission estate problem first and an adoption program second.
Cloud cost optimization. The bill keeps rising and nobody can explain it. A dashboard does not turn anything off. Read this if finance has asked for a forecast and engineering answered that it depends, because the fix starts with ownership enforced where resources are created.
Zero trust. Sound architecture, terrible program name. Sold as a transformation it becomes a twenty-four month roadmap where nothing gets safer for two quarters. Read this if an insurer, a client or an incident has set a date, and you need the controls that remove the most risk to land first.
Data platform modernization. Two teams, two numbers, one meeting, and a discussion about the data instead of the decision. Read this if reporting and AI are reading from different places, because the answer is one governed layer that both of them read from.
VMware exit. The renewal quote changed the conversation. Read this if you have a date, because the work that decides whether the move succeeds is a dependency map built from observed traffic and a rollback path per wave, not a lift-and-shift inventory in a spreadsheet.
The product decision comes last, and in writing
Copilot Studio or Azure AI Foundry. Microsoft Fabric or a semantic model on what you already have. Azure Arc on the existing server or a rehost. Each of those has a real answer that depends on where the data sits, what the workload needs from the network, and how much control you want over orchestration and evaluation. We make the call inside the first week of an engagement and record the reason, so the decision can be revisited by someone who was not in the room.
What we will not do is let the platform get chosen before the problem is described. When the tool is picked in the first meeting, the architecture that follows is a record of who ran that meeting rather than of what the workload needs.
One boundary is worth stating here. Five of these six are engineering problems, and they are Entrazure work: agents, Copilot, the data platform, the bill, the exit off VMware. Zero trust is the page that regularly touches both sides, because the same controls are sometimes bought to stop an attacker and sometimes bought to satisfy a regulator. When the driver is residency, jurisdiction, key custody or how quickly you could leave a platform, that is sovereignty work and it runs as Novarque by Entrazure, with the framework detail on the compliance pages.
What all six end with
Everything as code in your Git organization from the first commit, so ending an engagement early still leaves you holding what was produced. A written decision record with a reason against each choice. A named owner on your side, because a system nobody owns degrades on a predictable schedule. And a number: a cost model, a measured accuracy, a baseline, depending on what the problem was.
We do not open with a retainer. Each of these is delivered through a fixed-scope engagement, and the operating model after handover is a separate decision you make once you have seen the work, not a condition of starting it.
Which one to read first
If you can name the symptom, read that page. If two of them sound equally true, the constraint is probably underneath both, and the two-week assessment exists to find out which. If your sector is the reason the problem is hard, the industry pages carry that version of it, and the underlying capability is described on the services pages.
Or describe the meeting you just came out of and we will tell you which of the six it is, including when the answer is none of them.
Questions we get
How is a page here different from a service page?
A service page describes what we do and names the technology. A page here starts from what is wrong and often spans several services before it gets to a product name. The cost page touches governance, tagging, licensing and architecture; the agent page touches identity, search, networking and evaluation. If you already know which service you want, go straight to it. If you know the symptom, start here.
Can we run two of these at once?
Sometimes, and two pairs go together naturally: cost work rides along with a VMware exit because both need the same inventory, and Copilot work belongs next to zero trust because both are permission problems. Pairing an AI agent with a data platform rebuild is the combination we usually refuse, because the second one has to finish before the first can be trusted.
What if the real problem turns out to be none of the six?
That happens often enough that we built an engagement for it. The two-week assessment looks at the estate you actually have, ranks what it finds by severity against effort, and tells you which problem to buy first. It is read-only, it ends in a written opinion, and you keep the report whether or not we do any of the work in it.