Most Azure estates aren't waiting for migration. They're waiting for the second project.

A useful exercise, if you want to understand the current state of the UK enterprise cloud market, is to ask senior Azure consultants how many net-new migrations they did in 2025. The answers vary, but they cluster lower than people outside the industry expect. The big migration wave is mostly over. Most organisations of any size are already in Azure. The lift-and-shift work that defined 2020 through 2023 has largely been done.

What hasn’t been done, in most cases, is the second project.

The second project is the one nobody put on the original business case, partly because it didn’t have a nice name and partly because it’s the kind of work that doesn’t generate a press release. It’s what you do when your Azure estate is two or three years old, the original architects have moved on, the cost is somewhere between forty and eighty percent higher than the migration deck promised, and the senior IT team has the uneasy feeling that the platform has stopped working for them and started working against them.

This is the post-migration plateau, and it’s where most of the actual work is happening in 2026.

The symptoms are recognisable. The bill keeps creeping up and nobody can fully explain why. There’s a virtual machine running an application that two people left the company over and nobody is quite sure what it does, but turning it off feels risky. The number of Azure subscriptions has expanded from the original four to something like sixteen, several of which were spun up for projects that finished in 2023. There’s a tag policy on paper but compliance with it is somewhere around forty percent. Identity has sprawled into a thicket of service principals, managed identities, and legacy accounts whose lineage nobody can trace. The dev team wants to use a new service. The security team wants a new tool. The finance team wants to know why the AKS cluster cost twelve thousand pounds last month.

None of this is anyone’s fault. It’s the natural entropy of any complex platform left running for three years without the kind of structural review nobody budgets for in advance. And it’s the work that Azure Expert MSPs running modernisation programmes are now spending most of their time on, even though the marketing copy still leads on migration.

The shape of a good modernisation engagement looks nothing like the shape of a migration. A migration has a clear before-state and after-state, a defined scope, and an end. A modernisation engagement has none of those. It starts with what’s effectively a forensic audit: what’s actually running, what’s it costing, who’s using it, what depends on what, and which of the architectural decisions made three years ago no longer make sense for the way the business runs now. The output of the audit is rarely a single project. It’s usually a prioritised list of fifteen to twenty smaller pieces of work that, taken together, restore the estate to something resembling its original promise.

The pieces are unglamorous. Right-sizing VMs that were provisioned for peak load that never materialised. Decommissioning workloads that have outlived their purpose. Replacing IaaS with PaaS where the migration team didn’t have time to refactor. Consolidating subscriptions. Implementing the tagging policy that’s been sitting in a document since 2022. Rebuilding cost allocation so finance can actually trace spend to a business unit. Putting the security baseline in place that should have been there from day one but wasn’t, because the original goal was just to get out of the data centre by the deadline.

None of this is interesting in isolation. Cumulatively, it tends to take twenty to forty percent off the monthly Azure bill while making the platform meaningfully easier to operate, and it’s the kind of project where the consultancy genuinely earns its fee, because the customer’s internal team usually can’t do this work themselves. Not because they lack skill, but because they’re too close to the estate to see it clearly and too busy keeping it running to step back.

There’s a question worth asking at this point in the cycle. If your organisation moved to Azure between 2020 and 2023, when was the last time anyone looked at the estate end-to-end with fresh eyes? If the answer is never, the gap between what you’re paying and what you should be paying is almost certainly large enough to fund the review three or four times over.

The migration is done. The second project is overdue.