Cloud optimisation: Why migration is only the beginning
Leejay Bell, Senior Data Engineer at PBT Group
For many organisations, cloud migration is still treated as the major milestone. Once workloads have been moved, systems are live, and users can access what they need, the job appears complete. In many cases, that is when the real work starts.
Cloud does not automatically make a business more efficient. It gives the organisation greater flexibility, scalability, data availability, and technical capability, but those benefits can only be realised if the environment is continuously reviewed and improved. Without that discipline, the same inefficiencies that existed on-premises can simply be carried into a more dynamic environment, where costs estimated before migration can quickly become a more expensive operational reality.
This is why the cloud conversation has to move beyond “lift and shift”. Existing workloads cannot usually be moved from one platform to another without being adapted to the strengths, cost model, and operating requirements of the new environment. The more useful question is not only whether the workload can run in the cloud, but whether it should still run in the same way, whether the process design still makes sense, and whether the organisation is using the cloud environment in a way that supports performance and cost control.
On-premises environments often encouraged a different operating mindset. Infrastructure was already committed so that services could run continuously, processes could poll for work, and resources were used because they were already there. In the cloud, that behaviour has a cost implication. Services that run unnecessarily, workloads that are not properly calculated upfront, and processes that remain permanently active can quietly increase spend.
Understanding the work
Optimisation, therefore, starts with understanding the work before committing resources to it. Rather than relying on cloud infrastructure to determine what needs to happen, organisations should be clear about the workload, execution requirements, and expected outcome before services are triggered. This improves performance and helps control costs because cloud consumption is aligned with business needs.
This discipline becomes even more important after migration. Cloud environments can grow quickly. New services are easy to spin up, different teams can start experiments, and development work can spread across the environment faster than many organisations expect. Without active housekeeping, services can be left in limbo, unused resources can remain active, and costs can accumulate in places that are no longer visible to the business.
Reducing cloud sprawl is not only a technical clean-up exercise. Organisations must also have a clear administration and governance layer that continuously reviews what exists, what is being used, what should be decommissioned, and what is creating unnecessary cost or complexity. As cloud estates expand, this cannot be left to individuals who are already focused on delivery. It needs ownership, rhythm, and accountability.
Keep on reviewing
A practical optimisation approach should include regular review cycles after migration. This is when organisations can assess which workloads need to be modified, where processes should be redesigned, what skills are required, and which parts of the environment should be prioritised. Optimisation works best when it is incremental, with teams tackling specific areas in a structured way rather than trying to fix everything at once.
Skills are just as important as architecture. A cloud environment cannot be sustainable if the people expected to use, support, and maintain it have not been brought along. Many organisations have strong existing skills in legacy environments, such as SQL-based systems, but newer cloud approaches may require different tools, languages, and ways of working.
If teams are not trained and the rationale for change is not understood, adoption can stall. A technically sound solution can become dormant if nobody takes ownership of it or works through how it should be used. The result is often duplicated effort, with another customised solution later developed to solve the same or a similar problem.
AI complexity
The rise of AI adds another layer to this challenge. AI can help teams explore options, accelerate development, and generate ideas, but without coherent design, it can also expand scope and create technical debt. If teams accept AI-generated outputs without understanding the assumptions behind them, they may end up with over-engineered solutions that nobody understands well enough to support, refine, or correct. This is where accuracy and efficiency start to suffer.
The value of the cloud is not secured at the point of migration. It is protected through continuous optimisation, disciplined governance, cost awareness, skills transfer, and regular improvement. Migration may open the door to a more flexible and data-rich environment, but the long-term business case depends on what happens after the move.
