Locate the break in the workflow
Follow a real task from start to finish. Is the problem a missing screen, duplicate entry, an unclear decision or a system that cannot support the work? A new application will not solve a policy question, and an integration will not repair bad source information.
List what already works and the cost of changing it. Include user habits, data, permissions and dependencies. Protecting a reliable part of the process can be more valuable than making every part look new.
Configure or connect when the foundations fit
Configuration is worth examining when the product already supports the task but has the wrong fields, routes or permissions. Integration is worth examining when suitable systems need to exchange information or trigger an agreed action.
Check actual product editions, API permissions, record identifiers and usage limits. Define which system owns each record and what happens when an update fails. A connection needs an error owner and a recovery route, as well as the successful path.
Extend when a focused gap remains
An extension or small custom application can be useful when an important task falls outside the existing tools. Describe the missing job, who uses it and the information it needs. Keep the boundary clear so a focused tool does not quietly become a replacement programme.
For example, a request-review screen might sit above an existing customer record system. The brief should explain whether it reads, writes or only links to records. The example illustrates a design option; feasibility depends on the systems and access involved.
Replace only with a migration and operating plan
Replacement becomes a serious option when the current system cannot meet an essential requirement, is no longer supportable or creates unacceptable constraints. Compare migration effort, retraining, integrations, cutover and ongoing ownership alongside build or licence costs.
Ask each proposal to show assumptions, exclusions, acceptance checks and what happens if the transition needs to stop. Choose the smallest option that meets the actual requirement and can be maintained. A review may end with a configuration change rather than a software build.