--- title: The Autonomy Ladder | daily.dev description: A framework for governing AI agent autonomy by assigning action classes to one of six rungs rather than treating autonomy as a property of the model itself.... canonical: https://daily.dev/posts/the-autonomy-ladder-i2o9fwwmm twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: The Autonomy Ladder | daily.dev og:description: A framework for governing AI agent autonomy by assigning action classes to one of six rungs rather than treating autonomy as a property of the model itself.... og:url: https://daily.dev/posts/the-autonomy-ladder-i2o9fwwmm og:image: https://api.daily.dev/og/posts/I2O9fwwMm.png og:image:alt: The Autonomy Ladder og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # The Autonomy Ladder **[Medium](https://daily.dev/sources/medium_js)** · 26 min read · 1 upvotes · 0 comments ## Summary A framework for governing AI agent autonomy by assigning action classes to one of six rungs rather than treating autonomy as a property of the model itself. The core argument: the same model can be harmless or dangerous depending solely on what tools and credentials it holds. A six-rung ladder grades autonomy per action class, with the critical boundary between rung 3 (enumerated action space) and rung 4 (runtime-composed action space). Four axes — reversibility, stakes, observability, and accountability — determine the highest defensible rung, with observability scored first as the load-bearing link. Three failure patterns are named: marketing-driven rung inflation, approval-fatigue collapse, and covert irreversible actions hidden among reversible ones. A practical 'Rung Standing Test' (action class, rung number, recovery path) is offered as a pre-ship checklist. The piece also connects this framework to emerging regulatory requirements in Canada and the US AI RMF. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is the difference between rung 3 and rung 4 autonomy for an AI agent? Rung 3 means the action space is enumerated in advance, so the agent picks only from a fixed allow-list with caps and restrictions, and review happens mostly after the fact. Rung 4 means the action space is composed at runtime: the agent decides which actions, tools, and sub-goals are needed to satisfy a broader goal-shaped objective, and humans review the whole trajectory rather than individual actions. _Teams weighing how much autonomy to grant an agent can track this kind of framework thinking on daily.dev._ ### What four factors determine what autonomy rung an AI agent action should be granted? The four axes are reversibility, stakes, observability and accountability, scored per action class, with the grant set at the highest rung where all four pass. Reversibility asks if there's a tested recovery path; stakes asks if the worst outcome stays within an authorized loss; observability asks if a wrong action produces a signal in time; accountability asks if a named person can halt the action. Observability is scored first because reversibility and accountability both depend on it. _Anyone designing agent permission systems can follow this kind of production AI risk guidance on daily.dev._ ### Why do Gartner predict over 40 percent of agentic AI projects will be cancelled by 2027? Gartner's June 2025 forecast cites escalating cost, unclear business value, and inadequate risk controls as the reasons over 40 percent of agentic AI projects are expected to be cancelled by the end of 2027. This is framed as a forecast rather than a completed outcome, reflecting a mismatch between the autonomy level a system is sold at, designed at, and actually operated at in production. _Developers evaluating agentic AI investments can weigh forecasts like this via daily.dev before committing budget._ ## Similar posts on daily.dev - [The Oversight Ladder: the five levels of AI coding autonomy](https://daily.dev/posts/the-oversight-ladder-the-five-levels-of-ai-coding-autonomy-ds0q5sfsf) · Medium · 0 upvotes · 0 comments - [Agentic Autonomy Levels](https://daily.dev/posts/agentic-autonomy-levels-aklc8folj) · Addy Osmani · 4 upvotes · 1 comments - [What Is AI Autonomy? Levels, Meaning, and Credential Risk](https://daily.dev/posts/what-is-ai-autonomy-levels-meaning-and-credential-risk-au2tla5sg) · GitGuardian · 0 upvotes · 0 comments - [Agentic AI Security: Risks, Controls & Framework](https://daily.dev/posts/agentic-ai-security-risks-controls-framework-richrgmlz) · Orca Security Blog · 1 upvotes · 0 comments - [Towards a Self-Improving Autonomous Data Platform](https://daily.dev/posts/towards-a-self-improving-autonomous-data-platform-t1lvqucxn) · dltHub · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#ai-safety](https://daily.dev/tags/ai-safety), [#agentic-ai](https://daily.dev/tags/agentic-ai) [View this post on daily.dev](https://daily.dev/posts/the-autonomy-ladder-i2o9fwwmm) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"The Autonomy Ladder","url":"https://daily.dev/posts/the-autonomy-ladder-i2o9fwwmm","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/the-autonomy-ladder-i2o9fwwmm"},"datePublished":"2026-07-19T19:30:42.645Z","dateModified":"2026-09-14T08:02:41.324Z","description":"A framework for governing AI agent autonomy by assigning action classes to one of six rungs rather than treating autonomy as a property of the model itself....","image":"https://media.daily.dev/image/upload/s--1KxV4ohY--/f_auto/v1722860400/public/Placeholder%2007","thumbnailUrl":"https://media.daily.dev/image/upload/s--1KxV4ohY--/f_auto/v1722860400/public/Placeholder%2007","isAccessibleForFree":true,"articleSection":"Medium","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Medium","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/medium","url":"https://daily.dev/sources/medium_js"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/the-autonomy-ladder-i2o9fwwmm","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,ai-safety,agentic-ai","timeRequired":"PT26M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Medium","item":"https://daily.dev/sources/medium_js"},{"@type":"ListItem","position":3,"name":"The Autonomy Ladder"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/the-autonomy-ladder-i2o9fwwmm#faq","mainEntity":[{"@type":"Question","name":"What is the difference between rung 3 and rung 4 autonomy for an AI agent?","acceptedAnswer":{"@type":"Answer","text":"Rung 3 means the action space is enumerated in advance, so the agent picks only from a fixed allow-list with caps and restrictions, and review happens mostly after the fact. Rung 4 means the action space is composed at runtime: the agent decides which actions, tools, and sub-goals are needed to satisfy a broader goal-shaped objective, and humans review the whole trajectory rather than individual actions. Teams weighing how much autonomy to grant an agent can track this kind of framework thinking on daily.dev."}},{"@type":"Question","name":"What four factors determine what autonomy rung an AI agent action should be granted?","acceptedAnswer":{"@type":"Answer","text":"The four axes are reversibility, stakes, observability and accountability, scored per action class, with the grant set at the highest rung where all four pass. Reversibility asks if there's a tested recovery path; stakes asks if the worst outcome stays within an authorized loss; observability asks if a wrong action produces a signal in time; accountability asks if a named person can halt the action. Observability is scored first because reversibility and accountability both depend on it. Anyone designing agent permission systems can follow this kind of production AI risk guidance on daily.dev."}},{"@type":"Question","name":"Why do Gartner predict over 40 percent of agentic AI projects will be cancelled by 2027?","acceptedAnswer":{"@type":"Answer","text":"Gartner's June 2025 forecast cites escalating cost, unclear business value, and inadequate risk controls as the reasons over 40 percent of agentic AI projects are expected to be cancelled by the end of 2027. This is framed as a forecast rather than a completed outcome, reflecting a mismatch between the autonomy level a system is sold at, designed at, and actually operated at in production. Developers evaluating agentic AI investments can weigh forecasts like this via daily.dev before committing budget."}}]} ```