Microsoft Fabric User Group Pune

#### Description

Mode: Online
Date & Time: 26 Sept 10 AM IST
Managing time and attendance across a global workforce requires organizations to balance regulatory compliance, corporate policies, and operational efficiency. Traditional approaches often rely on manual reviews and fragmented processes, making it difficult to consistently identify compliance risks and potentially fraudulent activities in a timely manner.

This session presents a real-world success story of a Machine Learning-powered Working Hours Alert System (WHAS) for modernizing workforce compliance monitoring and fraud detection. The solution analyzes time and attendance data to identify compliance risks, detect anomalous working-hour patterns that may indicate fraud, and recommend schedule adjustments aligned with labor laws and organizational policies.

The session focuses on the data and machine learning aspects of building an intelligent compliance monitoring solution. Attendees will learn how workforce data can be analyzed to identify meaningful patterns, design effective alerts, support anomaly detection, and enable proactive intervention. The discussion will also cover implementation considerations, transparency, human oversight, and lessons learned from deploying an AI-driven compliance solution in a global enterprise environment.

By connecting machine learning with enterprise workforce data, the presentation demonstrates how organizations can reduce manual effort, improve consistency in compliance monitoring, strengthen governance, and support more informed operational decisions. Attendees will gain practical perspectives they can apply when designing scalable data and analytics solutions for workforce compliance and operational integrity.

#### Description

Agentic commerce is projected to redirect three to five trillion dollars in global retail spending by 2030, and AI-driven retail website traffic grew 693% year-over-year during the 2025 holiday season. Every major agentic commerce protocol today treats agent-initiated transactions as anonymous guest checkouts, which means loyalty tier, points, and benefits disappear the moment an AI agent completes a purchase on a customer's behalf. Since loyalty members generate two to five times the revenue of non-members, this is a costly blind spot, and it is also a data engineering problem Fabric is well suited to solve.

This session walks through a reference architecture for restoring tier recognition, benefit eligibility, and points accrual to machine-speed checkout flows operating under 100-millisecond latency budgets, built on Fabric's real-time stack. Eventstream and KQL Database power a Loyalty State Cache that serves pre-computed tier state in single-digit milliseconds using change-data-capture pipelines from the loyalty system of record. Data Activator drives a Benefit Eligibility Evaluator that resolves banner-parameterized offers and gated access in real time, while OneLake and Lakehouse provide the unified storage layer connecting transactional, promotional, and identity data without duplication. An asynchronous accrual engine settles points and benefit updates within seconds of payment confirmation, decoupled from the hot path.

The talk covers concrete implementation patterns: staleness policies tuned by benefit type, from 15-minute tolerances for standard discounts to under 2 minutes for gated event access; a layered security model using short-lived, single-use tokens; and graceful degradation logic that preserves the purchase with retroactive accrual when real-time loyalty resolution fails. Attendees will leave with a working blueprint for using Fabric's real-time intelligence and OneLake foundation to keep loyalty programs intact as an increasing share of retail transactions shift to autonomous AI agents.