Definition
AI workflow engineer
A developer who designs AI systems that produce code and other output.
From the episode “"The Agent Did It" Won't Hold Up in Court”Hear it on the podcast1 episode
Learn the language of AI-native founders
Every week: the new terms, tools, and workflows founders are using, straight from the podcast.
Related terms in Agents & Automation
AI PassengerA team member who defers to AI output and presents it as their own thinking without exercising independent judgment. Passengers say 'this is what AI thinks' rather than 'this is what I think, informed by AI.' In knowledge work, this is a liability — deferred thinking compounds until it fails visibly. Greg Shove's one-strike policy applies to this behavior.AI Thought PartnerUsing AI not to generate output, but to confirm decisions and surface blind spots before acting on them. A thought partner relationship involves genuine dialogue — asking AI to challenge your assumptions rather than validate them. Greg Shove identifies this as one of the most undervalued and underused applications of AI in business, especially for founders and executives.AI Voice AgentA software-based agent that conducts real phone conversations using natural language processing, text-to-speech, and contextual data from a CRM or database.AI workflowThe end-to-end process of building and running an AI system: preparing data, training, testing, deploying, and maintaining the model.Ambient agentsAI coding agents running continuously in the background on development tasks.Atomic TaskThe smallest unit of implementation work an AI coding agent can execute in a single turn without requiring additional context, clarification, or mid-task correction. An atomic task has a defined start state (what exists in the codebase), a defined end state (what should exist after), and clear acceptance criteria (tests pass, linting clean). The goal of requirement decomposition in AI-native development is to produce tasks atomic enough that the agent completes each one before context drift introduces errors. Tasks that require mid-execution judgment calls are not yet atomic — they need further decomposition.AutomationTechnology carrying out specific tasks according to rules a human has defined. Unlike an agent, it follows the instructions rather than deciding how to reach a goal.Autonomous sponsorship agentAn AI system that sources, matches, and closes sponsorship and partnership deals for creators without a human sales team in the loop.