AWS has put an AI agent in charge of investigating your cloud bill. The AWS FinOps Agent entered public preview on June 9, 2026, and it is free to try right now in us-east-1. If your current cost review process involves manual Cost Explorer queries, exported CSVs, and someone spending a Friday afternoon explaining why EC2 spend jumped 40% last week — this agent does most of that automatically.
What the Agent Actually Does #
AWS FinOps Agent connects to four native cost services: Cost Explorer (spend data), Cost Anomaly Detection (spike alerts), Cost Optimization Hub (rightsizing and idle resource recommendations), and Compute Optimizer (EC2, RDS, Lambda sizing). It also reads CloudTrail. That last part is the interesting one.
When a cost anomaly fires, most teams’ current workflow is: notice the spike on Monday, open a ticket, ask around, eventually find the EC2 fleet someone provisioned in us-west-2 last Thursday. AWS FinOps Agent collapses that investigation automatically. It detects the anomaly, cross-references CloudTrail events from the same window, identifies the infrastructure change that caused the spike, and produces an investigation summary with the likely root cause and responsible owner — then posts it to Slack or opens a Jira ticket.
The rightsizing workflow is similarly straightforward. The agent queries Cost Optimization Hub and Compute Optimizer, surfaces the top recommendations with estimated savings, and creates Jira tickets so the engineering team can pick them up without leaving their existing workflow. No more exporting CSV reports and pasting them into Slack.
How to Get Started (It Takes Four Clicks) #
Setup is faster than most AWS services:
-
Sign in to the AWS Management Console and switch to us-east-1(the agent only runs here, though it covers cost data from all regions) - Open the AWS FinOps Agent consoleand create your first agent - Complete the one-click IAM role setup — the role is read-only across cost services, with write access limited to EventBridge scheduling rules
-
Connect Jira (via the Forge app) and Slack (register the integration, bind a channel) The agent is free during preview with a monthly usage cap. This is the window to evaluate it before GA pricing arrives.
The Limitations Worth Knowing Before You Commit #
A few things to understand before you route your whole FinOps workflow through this:
AWS-only, full stop. No GCP, Azure, Snowflake, or AI provider spend. If you track OpenAI or Anthropic API costs alongside AWS, you still need a third-party tool. CloudZero, Finout, and nOps handle multi-cloud and AI provider spend — but they start at $6,000/year and up. For pure AWS shops, the native free option is a clear win.
It recommends; it does not act. The agent cannot terminate instances, modify resources, or execute cost-saving changes. It surfaces recommendations and creates tickets. A human or downstream automation executes. This is the right call for a v1 in preview — but if you were hoping for autonomous rightsizing, that is not what this is.
Slack sends a link, not inline data. When the agent posts to Slack, it delivers a link back to the FinOps Agent console rather than the full investigation summary inline. Users expecting a Slack-native experience where findings land directly in the channel will be surprised. It works; it is just less frictionless than the marketing suggests.
Account-to-owner mappings need governance. The agent uses context files that map AWS accounts to owners for anomaly routing. Stale mappings mean anomalies get sent to the wrong team, with the apparent authority of automation. Build a process to keep those current before you rely on them for incident response.
Why This Problem Is Worth Solving #
Cloud organizations waste roughly 27% of spend — over $100 billion globally in 2026. Idle compute accounts for 35% of that waste. Dev environments running around the clock when developers work eight-hour days represent a 67% theoretical waste rate on compute. Eighty-two percent of enterprises are already overshooting their cloud budgets.
The core problem is organizational: FinOps practitioners track the waste with dashboards that engineering teams rarely check. AWS FinOps Agent bets that routing cost intelligence into Jira and Slack — where engineers already live — is more effective than yet another cost dashboard. That is a reasonable bet.
The Bottom Line #
The anomaly investigation workflow is the standout feature. Automatically correlating a cost spike with CloudTrail events to identify the root cause and responsible owner is genuinely useful work that currently takes hours to do manually. For pure AWS teams with small FinOps functions, this agent likely pays for itself — which at zero cost during preview, is an easy bar to clear.
Multi-cloud teams should temper expectations: this does not replace a full FinOps platform. But for AWS-native organizations wanting to bring cost visibility into engineering workflows without paying for a third-party tool, the public preview is worth enabling today. Read the official launch post and the documentation for setup details.