AI and business intelligence work we have actually shipped

Situation

Flooring installs had conversion leakage and inconsistent customer experience. Existing measures didn't isolate what store teams could influence, and some stores routed around central services, increasing variability.

Task

Create operational KPIs reflecting controllable drivers of conversion and experience quality, align stakeholders, and operationalize the metrics for coaching and field action.

Action

Created an Estimate Quality KPI based on project square footage (rather than total cost) to normalize job scope and align performance to what associates control. Created a Store-Only Request Rate KPI to surface experience fragmentation. Aligned Merchants, Store Ops, leadership, Analytics, and Field teams on definitions. Published store dashboards with associate-level drill-down and field leadership reports.

Situation

Performance visibility across service teams was fragmented; leaders lacked consistent rollups while teams lacked role-specific KPIs tied to day-to-day execution.

Task

Create a scalable performance system combining shared north stars with team-controllable metrics, delivered in tools that work for both leaders and front-line coaching.

Action

Anchored north stars on Conversion Rate and Measure Volume. Built enterprise dashboards in Tableau and delivered individual report cards via Excel and Outlook automation for push distribution. Delivered role-specific KPI suites for field measurement (scheduling, on-time behavior, customer feedback, compliance) and call center teams (phone metrics, productivity, real-time execution).

Situation

Teams relied on proxy KPIs (signups and certificates) that didn't reflect real engagement, limiting content optimization and decision quality.

Task

Replace proxy measures with behavior-based engagement metrics and deliver dashboards content teams can use to improve performance.

Action

Built engagement measurement using verified participation signals with platform tracking in 8-minute intervals requiring explicit user confirmation. Standardized consumption metrics using minutes/hours watched and consumption rate, analyzed against customer consumption percentages per week. Delivered customized dashboards per content team enabling targeted actions: promote strong content, redesign, resequence, or retire low performers.

Situation

Teams lacked a structured experimentation process and consistent measurement methodology for product changes.

Task

Create a framework for reliable product testing and data-driven decision-making.

Action

Built an experimentation framework with standardized metrics, statistical rigor, and controlled rollout processes to enable teams to test product changes with confidence.

Situation

Demand for field services fluctuated by geography, and staffing decisions were reactive—capacity gaps appeared after demand spikes had already hit.

Task

Forecast service volume at granular geographic levels so hiring, training, and scheduling could ramp ahead of demand.

Action

Partnered with a data scientist to build and operationalize a forecasting model at the geographic quota-group level, incorporating seasonality, promotions/marketing effects, local trend signals, and operational factors. Translated forecasts into workforce planning inputs for hiring pipeline timing, training readiness, and scheduling capacity.

Situation

Customer survey outcomes and training data existed in separate systems across the U.S. dealer network, limiting the ability to tie learning programs to customer experience results.

Task

Unify survey outcomes with course completion history and produce actionable recommendations for training that improves specific deficiencies.

Action

Consolidated survey outcomes into a unified dataset spanning sales and service. Joined to dealer personnel training records. Built a recommendation model in R using classification and regression approaches including Random Forest, producing recommended courses by deficiency pattern. Delivered outputs in Tableau with individual employee drill-down and dealership rollups.

Situation

Avoidable Emergency Department visits were a key cost and care-quality issue. Care managers needed a consistent way to identify rising-risk patients and intervene earlier.

Task

Tier patient risk and recommend care intensity including visit frequency using clinical and non-medical context, operationalized for care manager workflows.

Action

Built risk tiering and visit-frequency recommendation models using classification and regression including Random Forest. Combined diagnostic codes with utilization and treatment signals, enriched with census and public data for contextual risk factors. Embedded outputs into call-center care manager workflows to steer patients toward primary care and virtual visits when appropriate.

Situation

Quarterly promotional games needed to drive participation and measurable behavior change, not just awareness—but no measurement framework existed to connect the funnel to actual outcomes.

Task

Design measurement and reporting tying the promo funnel (play → rewards → signup) to incremental customer behavior, and feed learnings into each quarter’s iteration.

Action

Defined KPIs across the full funnel plus behavior outcomes: participation, rewards earned and redemption, loyalty engagement, and visits per week. Integrated web/app activity, POS transactions, and loyalty usage into a cohesive measurement view. Compared promo vs. non-promo periods using visits-per-week to quantify lift, and segmented results to identify what drove stronger participation and frequency.

Situation

The institution implemented a real-time CDP and journey orchestration platform to enable unified profiles and personalized journeys, but banking-grade security constraints—especially encryption requirements—created architectural blockers.

Task

Drive implementation end-to-end: define measurable onboarding and cross-sell use cases, translate business goals into data and engineering requirements, and manage delivery risk when security requirements exceeded platform capabilities.

Action

Shaped priority use cases (onboarding into programs, cross-sell) with measurement expectations. Mapped data flows and defined requirements across stakeholders, analytics, and engineering. Managed delivery friction by adjusting sequencing and scope when vendor-promised features weren’t production-ready and when encryption requirements exceeded platform capabilities.

Situation

Reporting quality was poor and fragmented, with key information spread across multiple platforms and important data not consistently available. Existing BI reporting lacked reliability and trust.

Task

Build a central data foundation and deliver reporting stakeholders trust—then prove the new numbers are better than legacy reporting.

Action

Built a Microsoft Fabric Lakehouse as the central hub for data consolidation. Conducted discovery to locate sources, identify missing information, and understand team needs. Reconciled legacy vs. new logic, explained differences, and iterated until stakeholders accepted the improved definitions and outputs.

Situation

Analytics requests arrived through scattered emails and DMs to individuals, creating confusion, inconsistent requirements, and limited visibility into the backlog.

Task

Centralize intake and triage while improving request quality without forcing rigid forms on stakeholders.

Action

Implemented conversational intake using Microsoft Copilot Studio integrated with Jira, creating tickets that rolled into projects. Kept intake dialog-based to gather context naturally, with asynchronous follow-up for missing details instead of blocking submissions. Ran weekly triage meetings using the structured backlog to improve ability to pivot for urgent work without losing queue control.

Situation

Reports, queries, and dashboards were distributed across the organization with unclear ownership and duplicates, making replacement planning difficult and undermining trust in data.

Task

Inventory legacy reporting at scale, capture requirements consistently, identify duplicates, and drive consolidation toward a single source of truth.

Action

Built a conversational assistant that initiated discovery from forwarded emails containing existing reports, then interviewed stakeholders via chat to capture purpose, users, cadence, definitions, and desired future state. Persisted captured inventory into SharePoint and a structured repository in the Fabric Lakehouse. Focused on capture, organization, and duplicate detection across dozens of artifacts to support consolidation decisions.

Situation

Inconsistent metric definitions, uneven access practices, and limited auditability created reporting disputes and governance/compliance risk.

Task

Establish a governance operating model that standardizes definitions, enforces permissions, and provides audit trails.

Action

Operated a governance council and implemented board-approved governance policies. Defined metric standards, usage permissions, access levels, and audit logging. Enforced rules through platform controls: Fabric permissions, Power BI workspace and dataset controls, and Microsoft Purview for governance visibility and enforcement.

Situation

A factory fire destroyed the manufacturing facility, creating immediate operational and financial disruption requiring rapid documentation and stakeholder management.

Task

Own the financial documentation required for insurance and disaster recovery, and support continuity communications with vendors and customers.

Action

Led end-to-end financial documentation: coordinated inventory and asset records, assembled supporting evidence, and prepared structured loss documentation. Managed vendor and customer communications during disruption, including limited outsourcing while leadership evaluated viability.

Situation

Recurring operational costs tied to drug disposal were materially impacting spend with no existing initiative to address them.

Task

Deliver a targeted savings initiative for the CFO by reducing disposal cost without compromising compliance or operational needs.

Action

Executed a CFO-sponsored special project combining vendor renegotiation with process change: analyzed cost drivers, reframed requirements into negotiable terms, negotiated improved vendor pricing and terms, and supported implementation to ensure the change was repeatable.

Situation

Operational updates were scattered across teams and formats, slowing decision-making and obscuring performance visibility for leadership.

Task

Create a simple reporting system that gives leadership fast visibility into operational execution across maintenance and collections.

Action

Consolidated maintenance reports and tenant rent collection into a standardized spreadsheet with consistent structure and definitions, replacing fragmented updates and enabling quick review and follow-up.

Situation

Analytics adoption was constrained by uneven data literacy. Teams needed role-appropriate training and shared language for dashboards and KPIs.

Task

Improve analytics skills across the organization and create a repeatable learning cadence for data-adjacent staff.

Action

Ran bimonthly analytics roundtables for people working with data to learn and share practices covering dashboard literacy, KPI interpretation, experimentation concepts, platform basics, and responsible AI usage patterns. Delivered company-wide sessions showcasing delivered work and teaching fundamentals. Recorded all sessions to create a reusable training library.

Situation

Growing analytics backlog and inconsistent delivery were slowing business initiatives, with no clear operating model for the team.

Task

Build and scale the analytics organization to meet growing demand with repeatable processes.

Action

Expanded the analytics team with clear operating cadence, tooling standards, and governance. Introduced structured intake, prioritization, and delivery workflows to ensure consistent output quality.