AWS FinOps Agent (preview) features

AWS FinOps Agent listens for AWS Cost Anomaly Detection events and runs an investigation each time one arrives, so cost anomalies get caught while they are still small. To focus the agent on what matters most, you can include a filter in your automation prompt, for example, only investigating anomalies above a specific dollar threshold, so your team's attention stays on the highest-impact changes.

A Cost Anomaly Detection alert tells you something changed. AWS FinOps Agent takes the next step automatically: it correlates the cost change with AWS CloudTrail events, identifies the change that drove the anomaly, and produces an investigation summary with the likely root cause and the responsible owner.

AWS FinOps Agent can deliver the findings by opening a Jira ticket or posting to a Slack channel, so the engineer who owns the resource gets the context and can decide what to do next.

The agent runs in the background on the cadence you set, so anomaly investigations no longer rely on someone remembering to triage the queue. Findings are surfaced when there is something that warrants attention.

Lean FinOps team: A central FinOps team supporting many accounts uses event-triggered investigation to triage cost anomalies without anyone manually investigate each one. 
Hybrid FinOps model: A platform engineering team that owns shared AWS infrastructure routes each investigation to the engineer who owns the affected workload through Jira. 

AWS FinOps Agent lets engineers ask cost questions in natural language and get answers using your actual cost and usage data. An engineer can ask, "Why did my AWS cost go up last month?" and get a response that identifies the cost changes, the contributing services, and the underlying usage drivers.

To tailor the agent to your organization, you can upload context files such as account-to-owner mappings, team definitions, tagging conventions, and review cadences. The agent can leverage this context to interpret questions in your organization's terms, for example resolving "What's the cost of Team X?" to the specific accounts that team owns.

The agent remembers your preferences across sessions, and applies them in future runs. You can tell the agent to ignore a class of recommendation or apply a specific allocation rule, and it will respect that the next time it runs.

Create multiple chats, view and manage the task queue, approve ticket creation through a natural-language flow, and download generated reports from the Artifacts tab. Each task runs without blocking your next question.

Self-service for engineers: An engineer asks the agent why their service cost changed last week, and gets an answer with the cost movements and usage drivers, without filing a ticket with central FinOps.

Hybrid FinOps model: A FinOps team supporting hundreds of accounts redirects routine engineer questions to the agent, freeing time for chargeback design and leadership reporting.

Schedule recurring cost reports (e.g., daily, weekly, monthly) rendered in a downloadable, presentation-ready format in html, pdf, or ppt.

Pull recommendations from AWS Cost Optimization Hub and AWS Compute Optimizer, and summarize them into a Jira ticket so the engineering team can pick up the work in the tool they already use.

Create any recurring task in natural language, such as weekly reporting of cost optimization opportunities or a monthly cost outlier review, and the agent runs it on a schedule you define. 

Reporting at the cadence stakeholders need: A FinOps team replaces hand-built monthly executive reports and weekly business-unit chargeback updates with scheduled agent runs.

Optimization that reaches engineering: A FinOps team turns the monthly Cost Optimization Hub list into Jira tickets for the engineers who own each affected resource.

An agent can be created in any AWS account. Management account administrators can manage associated member accounts' cost and usage data through the AWS FinOps Agent, and member account owners can use AWS FinOps Agent to manage their own cost. 

Connect Jira space keys and Slack channels so the agent can create tickets and post messages in the workspaces your teams already use. Edit or remove connections from the Agent configuration at any time.