Selected strategy, systems, and operations work
AI strategy matters when the operating system underneath it works.
I connect the business problem, the measurable outcome, and the systems and agents that do the work. The examples below show how I have built AI-assisted workflows, improved client growth systems, and stabilized enterprise marketing infrastructure.
AI agent strategyMeasurementMarTech architectureClient transformation
Apps, agents, and product systems
Work I have shipped or am actively developing
The status is explicit: these projects range from an internal agent system used by the business to products still in development. Together, they show how I frame the problem, design the operating logic, and use AI-assisted tools to move an idea toward a useful system.
Shipped · internal
Competitive-intelligence agent system
A governed system of four role-aligned Custom GPTs that turned fragmented competitor monitoring into a consistent weekly decision signal for senior sales leaders.
My role
Problem framing, agent and prompt design, source governance, validation rules, operating cadence, and continuous refinement.
Approved sources→Agent monitoring→Pattern synthesis→Human review→Weekly brief
Four agents · roughly 20 approved sources · approximately 10 senior leadersFull case study ↓
In development · product system
Career Search OS / Project Pathfinder
An AI career agent designed to help people make better career decisions, not simply submit more applications. It turns durable career evidence into researched, role-specific application work while preserving human approval.
My role
Product vision, workflow architecture, source-library and memory design, automation planning, output standards, and beta definition.
Career evidence→Role discovery→Fit + research→Tailored assets→Human review
Designed for high automation with explicit review before submission
Early development · MVP architecture
Washington-first K–12 learning system
An AI-first alternative for children whose learning needs are not met by standard pacing. The model connects Washington requirements with competency-based progression and personalized, interest-led instruction.
My role
Product vision, learner and family requirements, standards-engine concept, AI-teacher operating model, MVP scope, and platform research.
Learner profile→Standards map→Personalized learning→Mastery checks→Progress record
Initially framed around real family needs, with broader K–12 application
Case study 01 · AI-native operating model
From ad hoc monitoring to a governed agent-assisted decision system
A multi-brand sales organization needed earlier, more consistent visibility into competitor moves, promotions, pricing, product changes, and tariff-related market signals.
The problem
Reactive intelligence
Monitoring depended on one-off searches and individual memory. Coverage varied, findings arrived late, and leadership had no shared weekly signal.
What I built
Four role-aligned agents
I designed four Custom GPTs aligned to sales groups and sister companies. Together they monitored an approved set of roughly 20 competitor and market sources, detected patterns, and drafted weekly briefs.
The control layer
Human judgment stayed accountable
Agents handled collection, comparison, and first drafts. I retained source validation, timing checks, pricing and tariff context, relevance decisions, prompt refinement, and final approval.
~10senior sales leaders received a consistent weekly briefing
The workflow replaced fragmented, reactive monitoring with a repeatable operating cadence and broader coverage. It also exposed the real constraint: incomplete source ownership and missing business context degrade agent output faster than the model itself.
Case study 02 · Measured client growth
Connecting demand generation, lifecycle, handoffs, and measurement
At The Pedowitz Group, I led marketing-operations and transformation work across enterprise client engagements. The work spanned discovery, automation, lifecycle, lead management, attribution, platform recommendations, implementation coordination, and ongoing optimization.
What I owned
Discovery, requirements, workflow and platform recommendations, implementation coordination, measurement design, optimization, and the connective tissue between Marketing and Sales.
How I measured it
Baselines and reporting tied channel behavior to the larger demand system. The goal was not a cleaner portal or more assets. It was a system that generated, nurtured, measured, and handed off demand more effectively.
How credit is assigned
These are representative outcomes produced by coordinated client and consulting teams. Client teams owned internal decisions and the close. My contribution was improving the connected operating system.
Case study 03 · Enterprise marketing infrastructure
Diagnose the system, protect live operations, and reset the growth curve
At Microsoft, recurring Marketo failures were being managed as separate incidents. I connected them to one structural issue: database growth was approaching an operating ceiling of roughly 10 million records. The work required an immediate continuity plan and a durable governance model.
The operating challengeMicrosoft
Database growth was outpacing the team's ability to remove records
The environment was adding roughly 1.2 million records per month while deleting only about 250,000. Treating each failure as an isolated ticket would have protected neither campaign delivery nor the business from an eventual capacity event.
- Elevated the shared root cause through support and leadership channels.
- Negotiated temporary Adobe capacity to protect active campaigns and buy time for remediation.
- Coordinated campaign operations, platform and technical teams, data and privacy partners, Adobe, director and VP leadership, with the SVP kept informed.
Records added monthly
Growth was compounding faster than the existing control process could absorb.
Records deleted monthly
The imbalance made the next failure predictable, even if its exact timing was not.
The hard decision
Reduce volume without creating campaign, compliance, or retention risk
The cleanup could not be a blanket purge. Records had to be classified by duplication, validity, inactivity, age, active campaign use, privacy requirements, and retention obligations before removal.
Remediate
Remove 2–4 million records safely
I led the coordinated cleanup of duplicate, invalid, inactive, and aging records, balancing the need for rapid relief with business, campaign, privacy, and retention requirements.
Govern
Make capacity an owned operating metric
We added automated deletion, stricter intake controls, retention rules, monitoring, alerts, and named ownership so database health could be managed before it became an incident.
Result
Stabilize the system and exit temporary capacity
The work prevented an outage, stabilized database growth, protected live campaign operations, and removed the need for the temporary capacity increase.
Case study 04 · CRM implementation and adoption
Make the operating process usable enough that teams choose to follow it
At Salem Group, the implementation risk was not the Salesforce software. It was moving Marketing and Sales onto a shared lead-management process without losing visibility or encouraging workarounds.
Salesforce user adoption in the first quarter
Adoption served as a practical system-health measure, showing whether the design worked in real user workflows rather than only in implementation documentation.
Design the work
Define the shared process before cutover
I designed the operating model around lead entry, ownership, handoffs, campaign representation, Sales visibility, and adoption monitoring.
Stay close to users
Use adoption friction as diagnostic evidence
During cutover, I stayed close to users so gaps could be corrected quickly and workarounds did not become the unofficial process.
Business effect
Improve the connective tissue between Marketing and Sales
The result was stronger lead management, clearer campaign execution, better Sales visibility, and a more reliable foundation for Marketing and Sales alignment.
How I work
Signal → System → Proof
AI does not remove the need for strategy. It makes unclear strategy and weak governance fail faster.
Frame the decision
Start with the business question, the evidence required, and the action the output needs to support.
Design the system
Map data, workflows, ownership, handoffs, measurement, and failure points before selecting the tool.
Assign the work
Delegate repeatable research and production to agents. Keep context, judgment, and approval with people.
Prove and improve
Measure business movement, inspect exceptions, refine the loop, and turn the result into a reusable playbook.
Current working toolkit
Custom GPTs · Claude · n8n · Marketo · Salesforce · Adobe Campaign · AEM · HubSpot · LeanData · analytics and attribution systems