The Same-Day Ops Agent Double Punch

On June 9, 2026, Datadog at DASH 2026 in New York announced a massive expansion of Bits AI: Detection, Investigation, Remediation, Infrastructure, Code, Release, Testing, Data Analysis, Chat, Memories, and Evals — over 10 agents and 100+ new features. The same day, AWS released FinOps Agent as a public preview, bundling Cost Explorer, Cost Anomaly Detection, Cost Optimization Hub, and Compute Optimizer into a Slack/Jira bot. AWS DevOps Agent had already gone GA in March.

Both companies are chasing the same prize: autonomous ops agents that detect anomalies, investigate root causes, and even write fixes. But their approaches differ fundamentally in data ownership, coverage, and positioning.

Data Source: The Structural Advantage

AWS owns cost data and CloudTrail as first-party data. FinOps Agent's anomaly investigation starts from a Cost Anomaly Detection event, correlates CloudTrail events, and identifies the IAM user or role behind the cost spike — all the way to a Jira ticket assigned to the responsible engineer. This flow is impossible for third-party FinOps tools because they ingest CUR with data lag and lack near-real-time CloudTrail access.

Datadog owns APM, logs, traces, RUM, and profilers. Bits Detection monitors real-time metrics plus historical baselines, service topology, and source-code context. Bits Investigation traces root causes, and Bits Code produces a production-ready PR on GitHub. Datadog's strength lies in performance and behavior anomalies; AWS excels at "money anomalies."

Coverage: Single Cloud vs Multi-Cloud

AWS FinOps Agent stays within AWS. It won't see Snowflake, Databricks, Google Cloud, or Vercel costs. For AWS-only shops, that's fine. For multi-cloud or SaaS-heavy orgs, Datadog's Bits AI covers everything, including BYOC Logs and the newly announced Federated Logs that search external stores like Databricks and ClickHouse from Log Explorer.

Positioning: Agents as Platform vs Agents as Monitored Objects

This is the core strategic difference. AWS treats AI agents as building blocks on Bedrock AgentCore, which provides Memory, Identity, Observability, Gateway, Runtime, and built-in tools like Browser Tool and Code Interpreter. DevOps Agent and FinOps Agent are standard agents on this layer. AWS wants you to run your agents on its platform.

Datadog positions itself as the observer of agents. The new Datadog Agent Console monitors all coding agents in an organization — Claude Code, Cursor, GitHub Copilot, and even Bits AI itself — in one UI. It answers "who used which agent, how much, and is it worth it?" AI Guard protects agents from prompt injection, tool misuse, and data exfiltration, regardless of vendor. Datadog's stance: "We'll give you visibility and governance over all agents."

The Turf War: Who Changes the Code?

Bits Code is Datadog's weapon. It takes signals from Error Tracking, APM Recommendations, Continuous Profiler, Test Optimization, Code Security, and Bits Investigation, then triages, locates code, writes a fix, runs tests, and opens the PR — all in one shot. AWS counters with Kiro, a spec-driven development environment, but Kiro is not yet directly integrated with DevOps Agent or FinOps Agent.

The tipping point is who completes the detection-to-fix loop inside their own UI. Datadog already does. AWS is building the runtime infrastructure but hasn't closed the loop for cost anomalies yet.

Practical Implications for SREs, FinOps, and Platform Engineers

If your org is 100% AWS, FinOps Agent for cost + DevOps Agent for incidents is a solid combo. If you run multi-cloud or SaaS-heavy workloads, Datadog Bits AI is unavoidable. Many orgs will split: AWS for cost, Datadog for performance.

But watch for lock-in. AWS's AgentCore is proprietary; Datadog's agent monitoring is vendor-agnostic. Choosing AWS for agents means betting on its execution environment. Choosing Datadog means betting on observability as the control plane.

What to Do Now

  1. Test AWS FinOps Agent in a sandbox AWS account. Let it ingest your cost data and see if the CloudTrail correlation actually identifies the right IAM user.
  2. Try Bits Code on a non-critical service. Feed it an APM recommendation and see if the generated PR is mergeable.
  3. Map your agent landscape with Datadog Agent Console. You might be surprised how many agents your team uses without central oversight.

The war for ops agents just started. Pick your side based on where your data lives, not on hype.