Automated vs Manual vs Consultant: Three Ways to Run an Azure Architecture Assessment
Automated tooling, the free Well-Architected Review, and a consultant. Three ways to assess an Azure estate, compared on speed, cost, depth and how long the answer stays true. Written by someone who sells one of them.
Marc Dekeyser |
Automated vs Manual vs Consultant: Three Ways to Run an Azure Architecture Assessment
There are three honest ways to find out whether your Azure estate is any good. Point a tool at it and get findings in hours. Fill in the free Microsoft Azure Well-Architected Review yourself. Or pay a consultant to come in and look.
Before you read another line: I sell the first one. So the sensible thing to do with this article is assume I am biased and check whether I earn my way out of it. I have tried to. If you want to skip to the part where I tell you not to buy the automated option, it is under “Picking one.”
People frame this as automation versus humans. Wrong question.
The one that matters is which method fits the decision in front of you, the budget you actually have, and how often the answer needs to be refreshed. Each of the three is genuinely best at one thing, and none of them is best at all three.
The three methods
Automated assessment points tooling at the environment, or at its definition, and generates findings against the Microsoft Azure Well-Architected Framework. Manual self-assessment is your own team working through a structured questionnaire, usually the free Well-Architected Review: roughly 60 questions across the five pillars. Consultant-led is a human, or a team of them, poking at the architecture, interviewing your people, and writing it up.
These are different outputs, not different grades of the same output. Automation gives you consistent, repeatable findings. The questionnaire gives you a self-scored snapshot. A consultant gives you interpreted recommendations shaped by your context.
None of them wins outright. They answer different questions.
How the three compare
Read the table, then read the paragraph under it. The paragraph is the part people skip and shouldn’t.
| Dimension | Automated assessment (e.g. PAA) | Manual self-assessment (Well-Architected Review) | Consultant-led engagement |
|---|---|---|---|
| Time to first output | Hours | Days (a workshop) | Weeks |
| Typical cost | Low, transparent (EUR 99–799/mo) | Free | High, contact-us (often five figures) |
| Currency over time | Continuous; re-runs cheaply | Stale once architecture changes | Point-in-time snapshot |
| Depth of judgment | Bounded by what tooling observes | Bounded by the team’s own knowledge | Deepest; contextual and strategic |
| Consistency | Deterministic, repeatable | Varies with who answers | Varies with the consultant |
| Compliance mapping | Per-finding to NIS2/DORA/ISO/SOC 2 | Not built in | Yes, if scoped (and priced) |
| Remediation output | Generated Terraform / Bicep | Guidance only | Tailored recommendations |
| Best at | Breadth, currency, drift detection | Free structured starting point | Context, strategy, hard trade-offs |
Here is the part the table can’t hold. A senior consultant is the only one of the three who can weigh business context, a planned acquisition, a regulatory deadline, how operationally mature your team really is, and then tell you which of ten findings actually matters this quarter. Automation can’t do that. It never will.
What automation does that no consultant will is run again next week, and the week after, at the same low cost, catching the drift that quietly turns every point-in-time assessment into a lie.
Picking one
Match the method to how often the answer has to be right, not just how deep you want it once. Three patterns cover most of what I see.
Automated assessment when you need findings fast, repeatedly, and mapped to compliance frameworks. Before a board update. During due diligence. On a platform that changes every sprint. It is quick and it stays current. It also only surfaces what tooling can observe, and it will never infer the strategy you never wrote down.
The manual Well-Architected Review when you want a free, structured way to start the conversation inside the team. Zero cost, first-party framing, good for getting everyone speaking the same language. The catch is that it is self-reported. A team that doesn’t know what it doesn’t know scores itself generously, and the score is stale by the next deploy.
Now the part where I argue against my own product. Do not buy the automated option if the thing you actually need is judgement. If your architecture is genuinely novel, if the trade-off in front of you is cost against resilience and somebody has to own the call, if the real blocker is organisational rather than technical, a consultant beats automation outright and it is not close. A tool can tell you that a subnet is open. It cannot tell you that the team who owns that subnet is being reorganised next quarter and will not action anything you send them. Buy the human.
That is not modesty. It is the honest boundary of what any tool, mine included, can see.
“Isn’t this just you selling your own product?”
Fair. Ask it, I would.
Here is the test I would apply if I were you. A vendor comparison is worth reading when it names the case where the vendor loses, and worth binning when every column quietly rolls back to “and that’s why you need us.” I’ve told you where the consultant wins and I’ve told you the questionnaire costs nothing. If you take one thing from this article and it is “run the free Well-Architected Review this month,” I’ll consider it a good outcome and you will have spent nothing.
Compete or complement?
They complement each other far more than they compete, because each one covers where the others are weakest. The pattern that works: run automated assessment continuously to hold the line on breadth and drift, use the free Well-Architected Review to get the team aligned on language and priorities, and bring in a senior consultant for the handful of decisions where judgment changes the outcome.
Automation makes that consultant worth more, incidentally. They spend their hours on interpretation and trade-offs instead of on inventory and box-ticking that the tooling already finished.
PAA is the automated one. I built it because I kept doing the same thing by hand: assess an estate, write it up, hand it over, and then watch the document rot while the estate moved on without it. Doing that inventory twice was tedious. Doing it a fourth time was insulting. So the tool does the boring, repeatable pass, and it stops exactly where judgement starts.
It’s read-only, for what it’s worth. It proposes changes. It does not make them.
FAQ
Is an automated Azure architecture assessment as good as a consultant? No, and it is not trying to be. Automation is faster, cheaper, and stays current, but a senior consultant brings contextual judgment and strategic interpretation that tooling cannot. They answer different questions. The strongest approach uses automation for breadth and currency and a consultant for high-stakes, context-heavy decisions.
Is the Microsoft Azure Well-Architected Review free? Yes. The Well-Architected Review is a free, first-party self-assessment of roughly 60 questions across the five pillars. Its weakness is not cost but reliability: it is self-reported, so its accuracy depends on the knowledge and honesty of the people answering, and it goes stale as the architecture changes.
How often should an architecture assessment be refreshed? As often as the architecture changes materially, which for an active platform is continuously. This is the core weakness of manual and consultant assessments: they are point-in-time. Automated assessment with drift detection is the only method that refreshes cheaply enough to stay current between major reviews.
When is a consultant the better buy than automation? When the decision is novel, high-stakes, or organisational. Re-platforming, a major architectural bet, or a trade-off between cost and resilience that someone has to call. Tooling reports state; a consultant reads context, politics, and intent, and none of those are observable from the control plane.
Can automated assessment map findings to NIS2 and DORA? Yes. Automated assessment tools can map each finding to control frameworks deterministically, meaning the same input yields the same mapping every time. Manual self-assessment does not include this, and consultant engagements include it only when scoped and priced for it. Deterministic mapping is what makes findings auditable.
Does automation produce remediation, not just findings? Some automated tools generate remediation as infrastructure-as-code, Terraform or Bicep you review and apply. This is a meaningful difference from the manual review, which gives guidance only. The code still needs human review before it is applied; automation proposes, it does not silently change your environment.
For a structured framing of what an assessment should cover, see the Azure architecture governance checklist, and for one pillar in depth, the Well-Architected security pillar walkthrough.
The short version
- Automated assessment produces findings in hours and stays current with continuous re-runs. Depth is capped by what tooling can observe.
- The Microsoft Azure Well-Architected Review is free and structured, but manual and self-reported. Its accuracy is only as good as the honesty and knowledge of whoever fills it in.
- A consultant brings the deepest contextual judgment and strategic interpretation, at the highest cost, as a point-in-time snapshot. On novel architecture and organisational context, they beat automation outright.
- Currency is where they split hardest. Automation re-runs cheaply; manual and consultant assessments go stale the moment the architecture moves.
- They complement each other. Automation holds breadth and drift, a senior architect holds judgment.
The right answer is rarely one method. It’s automation holding the line on breadth and currency, a free questionnaire aligning the team, and a senior architect spending scarce judgment where it actually moves the decision.
And if you only have budget for one of the three this quarter, pick the one that matches the decision you are actually facing, not the one that sounds most thorough.
