When should you use serverless for background jobs

lately, i've been thinking about offloading some of our longer-running background tasks, like report generation or data processing, from our main application servers. serverless functions (like AWS Lambda or Google Cloud Functions) seem like an obvious choice for this. but i'm curious about the breaking point. at what scale or complexity do you start seeing drawbacks compared to a dedicated queue + worker setup? are there specific types of jobs that are a clear 'yes' or 'no' for serverless in your experience