Older rule-based automation tools (drag-and-drop workflow builders with fixed triggers, conditions, and actions) require rebuilding rule logic whenever an automation's behavior needs to change. Current AI agents replace that rule logic with plain-language instructions: the agent's behavior — its triggers, ordering, and data access — is changed by editing a paragraph of English rather than rewiring a rules graph.
Practical implication: prefer building automations as natural-language agent instructions so that changing what triggers the automation, what order steps run in, or what data it touches is a quick text edit rather than a rebuild of deterministic workflow logic. This is the same shift underlying Specification-Driven Development, where the editable artifact is a spec document rather than code.
Nystrom argues natural-language agent workflows are cheaper to iterate on than older deterministic workflow-builder tools: changing behavior means rewriting plain-English instructions rather than reconfiguring branching rule logic, lowering the cost of each iteration and making it practical to tweak a workflow far more often.