--- title: AI Agents Are Non-Deterministic. So Are You. Deal with It. description: AI agents are non-deterministic, but so are human engineers — and we already solved that problem with guardrails like pull requests, policies, and pipelines.... canonical: https://daily.dev/posts/ai-agents-are-non-deterministic-so-are-you-deal-with-it--ig1cvpbvm twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI Agents Are Non-Deterministic. So Are You. Deal with It. | daily.dev og:description: AI agents are non-deterministic, but so are human engineers — and we already solved that problem with guardrails like pull requests, policies, and pipelines.... og:url: https://daily.dev/posts/ai-agents-are-non-deterministic-so-are-you-deal-with-it--ig1cvpbvm og:image: https://api.daily.dev/og/posts/Ig1CvPbvM.png og:image:alt: AI Agents Are Non-Deterministic. So Are You. Deal with It. 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. # AI Agents Are Non-Deterministic. So Are You. Deal with It. **[DevOps Toolkit](https://daily.dev/sources/devops-toolkit)** · 16 min read · 1 upvotes · 0 comments ## Summary AI agents are non-deterministic, but so are human engineers — and we already solved that problem with guardrails like pull requests, policies, and pipelines. The same approach applies to AI agents doing ops work. The proposed architecture uses Crossplane as a control plane to expose a constrained, compliant API instead of raw cloud APIs, GitOps (Argo CD/Flux) to ensure nothing reaches production without a traceable commit, and Kyverno policies to enforce hard rules automatically regardless of who or what made the change. This 'leash' squeezes non-determinism down to a single point (the declaration), while everything downstream becomes deterministic reconciliation. The stack is entirely CNCF open source — no new AI-specific platform required. Self-hosted model inference via a model plane is also discussed for organizations with data residency requirements. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How can I safely let AI agents provision cloud infrastructure without giving them direct access to AWS or GCP APIs? Route agent requests through a control plane like Crossplane instead of the raw cloud API, so agents declare desired state (like 'give me a cluster') rather than calling hundreds of cloud-specific API knobs directly. The platform team defines a constrained, compliant-by-construction abstraction, shrinking the blast radius to only what that internal API exposes. _Teams wiring agents into real infrastructure track patterns like this GitOps-guardrails approach on daily.dev._ ### How does GitOps combined with policy enforcement tools like Kyverno prevent bad changes from AI agents reaching production? Requiring all changes to go through Git, reconciled by Argo CD or Flux, removes any direct path from agents to the control plane, giving a full audit trail and revert capability. Kyverno then enforces non-negotiable policies before every change is applied, blocking violations regardless of whether the change came from an agent, a reviewed pull request, or a human at 3am. _Anyone hardening agent-driven deployments can follow this GitOps-plus-policy pattern via daily.dev._ ### When does it make sense to self-host AI model inference instead of using a hosted API? Self-hosting inference only pays off once an organization is large enough that the cost math favors owning GPU infrastructure, or when data cannot legally or contractually leave the building. For smaller shops, sending prompts to a hosted API remains cheaper and perfectly adequate; self-hosted inference can be run through the same control plane, like a 'model plane' built on Crossplane compositions, as regular infrastructure. _Teams weighing hosted APIs against self-hosted inference can compare trade-offs like these on daily.dev._ ## Similar posts on daily.dev - [Kubernetes Was the Easy Part](https://daily.dev/posts/kubernetes-was-the-easy-part-kzq9zt3ac) · Cloud Native Now · 1 upvotes · 0 comments - [Crossplane and AI: The case for API-first infrastructure](https://daily.dev/posts/crossplane-and-ai-the-case-for-api-first-infrastructure-8au3afsc9) · CNCF · 6 upvotes · 0 comments - [Why your DIY Kubernetes stack won’t survive the era of agentic AI](https://daily.dev/posts/why-your-diy-kubernetes-stack-won-t-survive-the-era-of-agentic-ai-mueojos06) · The New Stack · 2 upvotes · 0 comments --- Tags: [#kubernetes](https://daily.dev/tags/kubernetes), [#ai-agents](https://daily.dev/tags/ai-agents), [#gitops](https://daily.dev/tags/gitops), [#crossplane](https://daily.dev/tags/crossplane), [#kyverno](https://daily.dev/tags/kyverno) [View this post on daily.dev](https://daily.dev/posts/ai-agents-are-non-deterministic-so-are-you-deal-with-it--ig1cvpbvm) ```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":"AI Agents Are Non-Deterministic. So Are You. Deal with It.","url":"https://daily.dev/posts/ai-agents-are-non-deterministic-so-are-you-deal-with-it--ig1cvpbvm","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-agents-are-non-deterministic-so-are-you-deal-with-it--ig1cvpbvm"},"datePublished":"2026-08-03T16:16:32.589Z","dateModified":"2026-09-13T20:19:53.281Z","description":"AI agents are non-deterministic, but so are human engineers — and we already solved that problem with guardrails like pull requests, policies, and pipelines....","image":"https://i.ytimg.com/vi/G6J2Z2fhlUo/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/G6J2Z2fhlUo/sddefault.jpg","isAccessibleForFree":true,"articleSection":"DevOps Toolkit","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":"DevOps Toolkit","logo":"https://media.daily.dev/image/upload/s--VPsieX2o--/f_auto/v1704633572/logos/devops-toolkit.jpg","url":"https://daily.dev/sources/devops-toolkit"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-agents-are-non-deterministic-so-are-you-deal-with-it--ig1cvpbvm","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"kubernetes,ai-agents,gitops,crossplane,kyverno","timeRequired":"PT16M","video":{"@type":"VideoObject","name":"AI Agents Are Non-Deterministic. So Are You. Deal with It.","description":"AI agents are non-deterministic, but so are human engineers — and we already solved that problem with guardrails like pull requests, policies, and pipelines....","thumbnailUrl":"https://i.ytimg.com/vi/G6J2Z2fhlUo/sddefault.jpg","uploadDate":"2026-08-03T16:16:32.589Z","duration":"PT16M","url":"https://api.daily.dev/r/Ig1CvPbvM","embedUrl":"https://www.youtube.com/embed/G6J2Z2fhlUo"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"DevOps Toolkit","item":"https://daily.dev/sources/devops-toolkit"},{"@type":"ListItem","position":3,"name":"AI Agents Are Non-Deterministic. So Are You. Deal with It."}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-agents-are-non-deterministic-so-are-you-deal-with-it--ig1cvpbvm#faq","mainEntity":[{"@type":"Question","name":"How can I safely let AI agents provision cloud infrastructure without giving them direct access to AWS or GCP APIs?","acceptedAnswer":{"@type":"Answer","text":"Route agent requests through a control plane like Crossplane instead of the raw cloud API, so agents declare desired state (like 'give me a cluster') rather than calling hundreds of cloud-specific API knobs directly. The platform team defines a constrained, compliant-by-construction abstraction, shrinking the blast radius to only what that internal API exposes. Teams wiring agents into real infrastructure track patterns like this GitOps-guardrails approach on daily.dev."}},{"@type":"Question","name":"How does GitOps combined with policy enforcement tools like Kyverno prevent bad changes from AI agents reaching production?","acceptedAnswer":{"@type":"Answer","text":"Requiring all changes to go through Git, reconciled by Argo CD or Flux, removes any direct path from agents to the control plane, giving a full audit trail and revert capability. Kyverno then enforces non-negotiable policies before every change is applied, blocking violations regardless of whether the change came from an agent, a reviewed pull request, or a human at 3am. Anyone hardening agent-driven deployments can follow this GitOps-plus-policy pattern via daily.dev."}},{"@type":"Question","name":"When does it make sense to self-host AI model inference instead of using a hosted API?","acceptedAnswer":{"@type":"Answer","text":"Self-hosting inference only pays off once an organization is large enough that the cost math favors owning GPU infrastructure, or when data cannot legally or contractually leave the building. For smaller shops, sending prompts to a hosted API remains cheaper and perfectly adequate; self-hosted inference can be run through the same control plane, like a 'model plane' built on Crossplane compositions, as regular infrastructure. Teams weighing hosted APIs against self-hosted inference can compare trade-offs like these on daily.dev."}}]} ```