https://www.learnwithparam.com/newsletter One issue at a time on building AI systems that hold up in production: architecture, the open source that matters, and what to do on Monday.en https://www.learnwithparam.com/newsletter/p/who-decides-what-runs-nexthttps://www.learnwithparam.com/newsletter/p/who-decides-what-runs-nextThu, 08 Oct 2026 00:00:00 GMT<p>Every AI system design starts with one question: who decides what runs next?</p><p>Workflow, single agent or several agents. How to choose before you write code, and why Season 1 of Live Labs opens with it.</p><p>The question under every design review</p><p>When a team shows me an AI feature that is slow, expensive or unpredictable, the cause is rarely the model. It is a decision made in the first hour and never revisited: who decides what runs next.</p><p>If your code decides, you have a workflow. If the model decides, you have an agent. If one model hands work to other models, you have a multi-agent system. Each answer has a different cost, a different way to fail and a different way to debug.</p><p>Start with a workflow, and earn the agent</p><p>A workflow is a fixed sequence of steps where one or two of them call a model. The path is known, so you can test it, retry it and explain it to an auditor. Most business automation belongs here: triage an invoice, summarise a ticket, draft a reply for approval.</p><p>Move to an agent when the steps cannot be known in advance: the task needs the model to look, decide and look again. Research, debugging and open-ended support are good examples. The price is that every run can take a different path, so you need traces and evals before you need anything else.</p><p>A second agent must pay for itself</p><p>Splitting work across several agents feels like good engineering. Often it adds handoffs, duplicated context and more places for a run to go wrong. A second agent earns its place when the work splits into parts that need different tools or different permissions, or that can run in parallel.</p><p>A useful test: if you cannot write down what each agent owns and what it hands back, you have one agent with extra steps.</p><p>A short checklist for your next design review</p><p>Before anyone writes a prompt, answer these on one page.</p><p>Who decides the next step</p><p>Your code, the model, or a supervisor model. Write it down.</p><p>What happens when a step fails</p><p>Retry, ask a person, or stop. Each tool call needs an answer.</p><p>How you will see a bad run</p><p>Traces from the first day, and a dataset made from real failures.</p><p>What one request costs</p><p>Estimate tokens per run before you build, then measure it after.</p><p>Where a person approves</p><p>Anything that sends, pays or deletes waits for a human in the loop.</p><p>Why Season 1 opens with this</p><p>The first Live Lab, on Tuesday 20 October, builds one support-triage task three ways: as a workflow, as a single agent and as a supervisor with two workers. We trace all three, count the tokens and break each one on purpose.</p><p>The labs after it follow the same idea. Design first, then evals and tracing, durable workflows, an MCP gateway, ingestion and vector search at scale, browser agents, AI QA, a company brain and an AI SDR. Each one starts from a public GitHub repo and ends with something running on your own code.</p><p>If you take one thing from this issue: before your next AI feature, write down who decides what runs next. Reply and tell me what you wrote.</p>