Serverless computing — functions-as-a-service platforms like AWS Lambda, Azure Functions, and Google Cloud Run — is one of the most misunderstood architectural patterns in enterprise technology. It is simultaneously overhyped as a universal replacement for servers and underused by organisations that could benefit from it significantly.
The core proposition of serverless is simple: you write code, deploy it to the platform, and pay only for the compute time your code actually executes. You provision no servers, manage no operating systems, and worry about no scaling configuration. The platform handles all of that.
The workloads where serverless excels are event-driven: tasks triggered by events — an API call, a file upload, a database change, a schedule — that run to completion and do not require persistent state. Image processing, data transformation pipelines, webhook handlers, scheduled reports, notification dispatchers. These are serverless's natural domain.
The workloads where serverless struggles are those with high concurrency and sustained, constant load. A high-traffic API that receives ten thousand requests per second continuously will be more expensive on serverless than on a well-provisioned container cluster. The per-invocation pricing that makes serverless cheap for sporadic workloads makes it expensive for heavy, constant ones.
Cold start latency is the other key consideration. A serverless function that has not been called recently may take several hundred milliseconds to initialise before executing. For user-facing APIs where latency matters, this requires mitigation strategies: provisioned concurrency, warm-up pings, or simply choosing serverless for background work where latency is less critical.
For enterprises beginning to evaluate serverless, the recommendation is to start with internal, non-latency-sensitive workloads. ETL jobs, report generation, data validation pipelines, and scheduled maintenance tasks are ideal candidates. Build operational experience with the model before deploying it to customer-critical paths.