A logical unit of compute that does work, in Kubernetes specifically: Deployments, StatefulSets, DaemonSets, Jobs, CronJobs.
A workload is a higher-level abstraction over pods that describes how many replicas, what update strategy, what scheduling rules to use. In Kubernetes the canonical workload types are Deployment (stateless replicated), StatefulSet (stateful with stable identities), DaemonSet (one per node), Job (run-to-completion), and CronJob (scheduled job). Outside Kubernetes the term is generic: an ECS service, a Lambda function, a Cloud Run revision, a long-running VM, all of these are workloads.
Picking the right workload type is the most consequential Kubernetes design decision per workload. Running a stateful database as a Deployment (instead of StatefulSet) silently loses state on rescheduling. Running a job as a long-running Deployment instead of a Job leaves zombie pods behind. The defaults are wrong for many cases and the failure modes are subtle, so the workload-type choice deserves explicit review.
See the part of the platform that handles workload in production.