Darren GoonawardanaChief Executive Officer & Co-Founder, Collectiv · Chief Executive Officer & Co-Founder at Collectiv

Your AI Agents Need Job Descriptions

14 Oct 2026 · 1 min

Watch on YouTube

1:25

In this concise presentation, the speaker argues that AI agents should be treated as a managed workforce rather than ad‑hoc tools. He critiques the common approach of hiring AI agents by first selecting tools or prompts and then asking if they can perform a task, calling it backwards. Instead, he proposes that each AI agent be given a formal job description consisting of five components: a defined role, a measurable outcome, a decision boundary, an escalation rule, and a review cadence. This structure provides clarity for operating systems, turning AI agents from unpredictable "slot machines" into reliable parts of a company's operating model. The speaker concludes that CEOs who succeed with AI will be those who redesign work with clear, shared job definitions for both humans and machines, not merely those who accumulate more AI tools.

Chapters

If your AI agents don't have job descriptions, you don't have an AI workforce. You have automation with a name tag. Most companies are hiring AI agents the wrong way. They start with tools, they start with prompts, they start with, can this agent do the task? That's backwards. If you were hiring a human executive assistant, analyst, recruiter, or operations lead, you wouldn't just say, be useful. You define the role. What outcomes does this person own? What decisions can they make? What tools can they use? When should they escalate? What does good performance look like? AI agents need the same thing, not because we're pretending they're people, but because operating systems need clarity. An agent without a job description becomes a slot machine, sometimes useful, sometimes noisy, impossible to manage. An agent with a clear mandate becomes part of the company's operating model. That means every agent should have 5 things: a role, a measurable outcome, a decision boundary, an escalation rule, and a review cadence. That's how you move from we use AI to we manage an AI workforce. The CEOs who win with agents won't be the ones with the most tools. They'll be the ones who redesign work clearly enough that both humans and machines know what job they're supposed to do.