Transformation
Work
Building proprietary AI platforms, deploying enterprise AI programs, and architecting the infrastructure that connects them.
5
PLATFORMS BUILT100+
TEAM SCALED200%
YOY MRR GROWTHIRIS
AI Assistant & Agent Orchestrator
Power Digital's proprietary AI platform. Conversational interface connecting the entire data ecosystem — Gong, Salesforce, Google Drive, Nova, Signal, and Looker — via MCP integrations. IRIS is the OS; agents are the workers.
- —1,900+ messages in 30 days, doubling week-over-week with no formal launch
- —109 active users growing ~3x month-over-month
- —Automated deliverables: MTD reports, QBRs, client briefs, handoff docs, action item recaps
- —Slide deck generation with 3 output modes (Full AI, Hybrid, Templates)
- —Power Circuit Narrative automation for prospect qualification
- —Cold outbound email generation using SimilarWeb data for sales prospecting
- —Prompt routing guardrails blocking unsanctioned budget recommendations without MMM validation
- —Agent orchestration: SEO, email, content, landing page, and paid media agents route through IRIS
- —Tesseract (closed beta): multi-level reasoning layer feeding IRIS outputs
Signal
KPI Intelligence Platform
Real-time executive dashboard surfacing revenue per employee, client retention risk, profitability, hiring pipeline health, and capacity planning across the entire organization.
- —Direct Vena API integration — eliminated manual exports and stale spreadsheet workflows
- —Client profitability module surfacing margin data, preventing overhiring, identifying underperformers
- —CSM weekly reports with automated Slack delivery and health scoring across 500+ accounts
- —Division and department-level RPE (revenue per employee) drill-downs
- —Hidden Innovation section (leadership-only) that timestamps AI solutions and tracks metrics over time
- —Single base KPI model providing consistency across all dashboard pages
- —QA agent prototype for automated data validation
nova
Data Warehouse & Client Portal
Proprietary data warehouse and client-facing portal platform. Oversee the AI integration layer, product evolution, and strategic positioning as a knowledge graph across 500+ clients.
- —InsightsX agent: autonomous weekly analysis of client call notes → structured Slack digests
- —Nova Cloud as knowledge graph — single source of truth across 500+ clients
- —3-month warehouse refactoring: 600+ models simplified to incremental DBT jobs
- —Client portal adoption driven from 30% → target 80% through mandatory walkthroughs and UX improvements
- —Proposed extension: auto-creating Jira tickets from agent-identified improvements
Churn & Expansion Models
ML-Powered Client Prediction
Client risk prediction and service expansion models that shift the agency from reactive account management to predictive — teams see risk before clients signal it.
- —90-day client churn risk prediction using LightGBM and XGBoost
- —Live prediction pages deployed to Signal with direct Snowflake integration
- —Scores surface in IRIS for account teams during client conversations
- —Companion service expansion model predicting upsell candidates — validated and moved to production
- —Retention and expansion treated as two sides of the same intelligence
- —Scores flow directly to Salesforce for account team action
Claude Enterprise Rollout
Led Anthropic Claude deployment across the organization. 80%+ active utilization through organic, peer-led adoption — no formal training mandate.
- —Marketing Science: MMM automation compressing media planning from days to hours
- —Sales: brand assessments running 3-4x/week for prospects like Briogeo and Liquid IV
- —People Ops: automated exit trend analysis, pulse survey data, NPS processing
- —Account teams: lifecycle audits, client deck assembly, strategic recommendations
- —Web Dev & CRO: spread organically from Innovation without structured rollout
- —Cost governance layered in as usage grew — Opus and Cowork usage limits, cross-tool oversight
- —21 documented use cases across 6+ departments
Claude Code Engineering
Drove Claude Code adoption across the engineering organization as a production development tool, standardizing AI-assisted development workflows.
- —PHPUnit test automation: per-feature testing from ~2 hours to ~30 minutes
- —Production deployment via GitHub integration — eliminated manual release steps
- —Planning-first workflow: execution plans drafted before AI agent handoff
- —Context management best practices (/clear, /compact) standardized across team
- —Cursor + Claude Code integration formalized with workshops
- —Build-vs-buy architecture decisions drafted via Claude Code (e.g., Rollstack evaluation)
ChatGPT Deployment
Deployed alongside Claude as part of a unified, centralized LLM program with governance across all tools.
- —Ad creative mockups across multiple placements
- —Unstructured brief formatting and stakeholder brain dump structuring
- —Lifecycle audit deck production
- —Managed under centralized governance with Claude and Gemini
AI Agent Pipeline
Built internal infrastructure for developing, deploying, and measuring AI agents across the organization.
- —Paid Media Auditor: 6 audits/week, fully adopted with quantified ROI
- —Asana Agent Suite: automated triage, health monitoring, task routing on live boards
- —Asana agent explicitly tested against an open PM headcount — goal: eliminate backfill
- —Nova InsightsX Agent: autonomous weekly client call analysis → Slack digests
- —Client-facing agent team: 1 developer + 1 account director per client
- —Agent-as-a-service incubation for new revenue line via R&D
MCP Gateway Architecture
Centralized Model Context Protocol strategy connecting all internal AI systems. LightLLM architecture providing single-endpoint access control and security. Mandated small, reusable, single-responsibility MCP servers across all teams. Strategic position: build internal agent capabilities before connecting to external ecosystems.
LLM Governance Layer
Unified oversight managing Claude, ChatGPT, Gemini, and Cursor. IRIS prompt routing guardrails ensuring AI outputs stay within approved analytical frameworks. Seat audit automation tracking utilization across all tools and identifying unused capacity for cost optimization.
Impact Measurement Engine
Automated case study capture with structured metadata. Weekly and monthly rollups with theme analysis, tool adoption tracking, and trend identification. Auto-updating presentation deck for executive and board-level reporting. Conservative methodology: stated frequency and headcount only, no company-wide extrapolation.
SaaS Pricing & Monetization
Rebuilt Power Digital's pricing structure end-to-end through nova. Implemented SaaS payment frameworks and facilitated over 200% YOY MRR growth in the first three months of launching the monetized product. Directly connecting innovation initiatives to revenue outcomes.