{
  "id": 7318003,
  "title": "Designing agentic development workflows: what a workflow is, and what it is made of",
  "url": "https://urgent.news/2026/09/14/designing-agentic-development-workflows-what-a-workflow-is-and-what",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-14T12:45:42.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/dani_shemesh/designing-agentic-development-workflows-what-a-workflow-is-and-what-it-is-made-of-j20"
  },
  "original_language": "en",
  "account": "What the series is for\nGive a team an AI coding tool and you have not provided them with a structured way of working. Each engineer using the same agent will interact with it in different ways. Some explore the codebase before planning, others dive straight into coding. A third might write tests after, if at all. The same engineer could vary their approach on different days. The phases are experienced internally by individuals, reconstructed each time, and the tool is unaware of your team's review process, repository conventions, or the mistakes made in previous iterations. Attention can go awry in both directions. People may observe a model reasoning through a decision they never intended to challenge, then overlook the moment where their judgment mattered most, only to discover afterward that it took an unacceptable route. Creating a model to write code is straightforward. The challenging aspect is ensuring it consistently performs the same actions, stops at predetermined points, and leaves behind something that can be reviewed. This series aims to address how to design such functionality: a suite of agentic development workflows, their components, the rationale behind their structure, the benefits they provide, how to share them with others, and the limitations that still exist. It is derived from workflows we have created, built, and executed on real work. The content is not tied to a specific product or codebase; the assertions are about the patterns.\n\nPart Question it answers\n1. What a workflow is, and what it consists of: defining the concept and illustrating a single run.\n2. The principles that uphold the workflow's integrity: the underlying mechanisms and guiding principles.\n3. The costs of control: determining the scope of actions, the number of approvals required, the model to be used, and the overall value offered.\n4. Evaluating effectiveness and worth: assessing whether the output is correct, identifying process pain points, and measuring the overall impact.\n5. Shipping, porting, and running unattended: packaging, harness portability, and the path towards autonomous operation.\n6. Identifying limitations and initial steps: understanding the constraints and prioritizing the most critical areas for improvement.",
  "summary": "What the series is for Give a team an AI coding tool and you have not given them a way of working. Everyone gets the same agent and uses it differently: one engineer explores the codebase before planning, another goes straight to code, a third writes the tests afterwards if at all. The same engineer does it differently on a different day. The phases live in people's heads and get reconstructed…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}