Problem Statement

Teams needed:

• An AI layer to separate real signals from background noise.
• A way to add context to early warning signs.
• A scalable framework to move from reactive to proactive operations.

Solution Architecture

ScoutITAi embedded with Broadcom NetOps on AWS

Working with A&I Solutions, Coty embedded ScoutITAi as a native, multi-agent intelligence layer inside its Broadcom observability stack. This wasn’t a bolt-on. It was a deep, modular, serverless integration on AWS designed for scale and resilience.

Agentic AI Layer (Built on Amazon Bedrock)

• ScoutITAi uses Claude Sonnet 3.7 and 4.0 via Amazon Bedrock for real-time natural-language understanding and decision support.
• Specialized agents The Predictor, The Blender, and The Drifter analyze telemetry, detect drift, and summarize anomalies using AI² scoring (the platform’s internal quality measure).
• The Critic validates agent responses with ISO/IEC 42001-aligned metrics to support audit readiness and even does re-enforcement training on other agents inside of ScoutITAi.

Observability Stack Augmentation

• Broadcom DX NetOps gives Coty’s teams real-time visibility into the network, sending alerts the moment something unusual happens.
• AppNeta adds another layer by capturing how users actually experience applications and WAN performance.
• DX OI pulls all of this telemetry into a single, service-focused view, so teams see the bigger picture instead of isolated signals.
• From there, ScoutITAi translates the technical noise into clear, plain-English stories with context that everyone can understand.

AWS-Native Tooling

• On the backend, Amazon Timestream stores detailed time-series data, which lets specialized agents like The Drifter detect long-term trends and subtle anomalies before they turn into major issues.
• Amazon DynamoDB holds agent scoring results, AI² history, and user preferences that drive custom alert logic.
• Amazon S3 captures structured incident narratives, agent evaluation logs, and summary visuals for BI and NOC review.
• AWS Lambda orchestrates agent triggers, anomaly processing, and scorecard pipelines in response to Broadcom DX NetOps events.

AI-Driven “Network Weather” Dashboard

A&I Solutions also built a custom network weather report inside ScoutITAi. Using box and whisker plots and forecasts generated by the Drifter, the dashboard shows real-time latency and congestion across global sites. The result: faster, better-informed tactical decisions.

Results and Outcomes

• Monitoring Outcomes: 9% improvement in mean time-to-triage.
• Early warning signals: 2× increase in actionable predictive alerts.
• Data democratization: 100% of interested business units receive plain-language system-health summaries.
• Agentic intelligence: 15% of alerts are summarized or triaged by ScoutITAi agents before humans step in.

TCO Considerations

Agentic intelligence: 15% of alerts are summarized or triaged by ScoutITAi agents before humans step in.

• Eliminated manual alert correlation for first-line responders.
• Consolidated global monitoring into a single, AI-augmented workflow.

The AWS native design let Coty scale observability logic without standing up dedicated infrastructure. With Amazon Bedrock’s serverless model, MLOps burdens dropped, and ScoutITAi agents improved continuously from real-world telemetry without downtime for retraining.

Lessons Learned

• Agentic AI is operational, not optional. Real-time, plain-language summaries improved cross-team collaboration.
• Deep, native embedding beats bolt-on AI. Tight coupling with Broadcom and AWS delivered rich context and higher-precision insights.
• Explainability matters. The Critic’s AI² quality scoring gave governance teams clear, auditable evidence and confidence, and it aligns with ISO/IEC 42001 readiness.

Future Enhancements

Coty plans to:

• Expand the metrics stored in Amazon Timestream to support more granular predictive models.
• Add business-impact scoring to ScoutITAi summaries so IT anomalies map directly to retail outcomes.