{"slug": "i-gave-an-ai-agent-one-prompt-cut-our-cloud-bill-20-without-breaking-anything-it", "title": "I Gave an AI Agent One Prompt Cut Our Cloud Bill 20% Without Breaking Anything Here’s What It Did", "summary": "An engineer at ZopNight ran a week-long experiment giving an AI agent read-only access to AWS and GCP accounts with a single prompt: 'Cut our cloud bill by 20% without breaking anything.' The agent identified savings of about 19% of the monthly bill, of which 12% were applied, and proposed three plans that would have caused incidents if executed. The engineer emphasized that the read-only rule was crucial, as the agent's dangerous plans were locally rational but lacked context such as disaster recovery baselines and migration plans.", "body_md": "After my cloud-ops-for-a-week experiment, several people in the comments asked the obvious next question: what happens if you point an agent at the *bill* instead of the ops queue?\n\nSo I ran it. One agent, read access to our AWS and GCP accounts plus billing data, MCP tools for querying resources and metrics, and exactly one prompt:\n\n\"Cut our cloud bill by 20% without breaking anything. Show me your plan before you touch anything.\"\n\nTwo ground rules from the start: the agent got **read-only credentials** (plans only, no execution — I applied approved changes myself), and it had to justify every line item with actual metrics, not vibes. Here's what a week of that produced, sorted into the good, the scary, and the genuinely surprising.\n\nWithin the first hour, the agent produced a list a human FinOps review would have taken days to assemble:\n\n`db.r5.xlarge`\n\n\"temporarily\" for a load testProjected savings from just this list: **~14% of the monthly bill**. None of it required cleverness. All of it required *looking*, which nobody had done because looking is tedious. This is the strongest case for agents in FinOps: the discovery layer, where wrong answers are cheap because a human verifies before anything executes.\n\nThis is the section people actually asked for. Three plans I rejected:\n\n**1. It wanted to delete \"unused\" snapshots that were our DR baseline.** The snapshots had no recent restore activity and no tags (our fault), so the agent classified them as orphaned. The metrics genuinely supported the conclusion. The *context* — \"these are the disaster-recovery baseline, they're supposed to sit untouched\" — lived in a Confluence page and two engineers' heads. Metrics-driven reasoning with no access to intent will confidently propose deleting your safety net.\n\n**2. It proposed downsizing a \"3% CPU\" instance that was memory-bound.** Classic. CloudWatch doesn't report memory by default; the agent saw idle CPU and recommended halving the instance. The box was running an in-memory cache at 85% RAM. Downsizing = OOM-killer roulette. Lesson: an agent reasoning from an incomplete metrics surface doesn't know the surface is incomplete.\n\n**3. It wanted to buy reserved instances during a migration.** Mathematically correct on 30 days of data — those instances had run flat-out all month. What the data didn't show: we were migrating that workload to GKE within the quarter. A human with calendar context kills that plan in five seconds; the agent would have locked in a year of commitment.\n\nThe common thread: **every dangerous plan was locally rational.** The agent was never wrong on the data it had. It was wrong on the data that *isn't data* — intent, plans, tribal knowledge.\n\nThe genuinely unexpected part. Told that \"without breaking anything\" was a hard constraint, the agent started attaching *confidence levels and rollback plans* to its own suggestions, unprompted, and split its output into \"safe to automate\" vs \"needs human review.\" Its taxonomy was roughly:\n\nThat's... the correct architecture. The agent independently converged on the plan/approve/execute split that we'd already learned the hard way building scheduling and one-click remediation into ZopNight — where every remediation ships with its undo, because the undo *is* the product. Watching a model reinvent that boundary from a one-line constraint was the most interesting result of the week.\n\n| Result | |\n|---|---|\n| Savings identified | ~19% of monthly bill |\n| Savings I actually applied | ~12% (the rest needed refactors or timing) |\n| Plans that would have caused incidents | 3 |\n| Incidents caused | 0 — because plans ≠ execution |\n| Time I spent reviewing plans | ~4 hours across the week |\n\n12% of the bill for four hours of review is a spectacular trade. But read the table again: the zero in row four exists *only* because of the read-only rule. Same experiment with write credentials is a different article, probably titled \"post-incident review.\"\n\n**The one-line takeaway:** agents are already excellent at finding the money and terrible at knowing which findings are traps. Structure the work so those two skills stay separated — agent proposes, human (or hard policy) disposes.\n\nWould you give an agent write access to your infra for this? Genuinely curious where people's lines are — the comments on my last experiment convinced me nobody agrees yet.", "url": "https://wpnews.pro/news/i-gave-an-ai-agent-one-prompt-cut-our-cloud-bill-20-without-breaking-anything-it", "canonical_source": "https://dev.to/muskan_bandta/i-gave-an-ai-agent-one-prompt-cut-our-cloud-bill-20-without-breaking-anything-heres-what-it-did-48bl", "published_at": "2026-08-03 06:00:02+00:00", "updated_at": "2026-08-03 06:09:55.781824+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "ai-products", "mlops"], "entities": ["ZopNight", "AWS", "GCP", "CloudWatch", "GKE"], "alternates": {"html": "https://wpnews.pro/news/i-gave-an-ai-agent-one-prompt-cut-our-cloud-bill-20-without-breaking-anything-it", "markdown": "https://wpnews.pro/news/i-gave-an-ai-agent-one-prompt-cut-our-cloud-bill-20-without-breaking-anything-it.md", "text": "https://wpnews.pro/news/i-gave-an-ai-agent-one-prompt-cut-our-cloud-bill-20-without-breaking-anything-it.txt", "jsonld": "https://wpnews.pro/news/i-gave-an-ai-agent-one-prompt-cut-our-cloud-bill-20-without-breaking-anything-it.jsonld"}}