AI, applied
not performed.

The foundation

My process, built over twenty years.

I solve for friction. I don't make assumptions. I diagnose before I prescribe. I move toward meaningful goals instead of just moving. That's the foundation. Everything else -- including AI -- works because of that foundation.

The process doesn't change. What AI does is let me move through it faster and more consistently.

I start by understanding what's actually broken, not what someone thinks is broken. I ask questions. I listen for the gap between what people are asking for and what they actually need. I build something that solves the real problem, then I maintain it because I know it works.

That's the framework. It's worked across 20+ years and dozens of organizations. Different tools, different platforms, same core approach.

AI fits into this process the same way every other tool does: it accelerates parts of the work that don't require judgment, so I can spend time on the parts that do.

The inventory

AI Skills

Tool Proficiency & Selection

• Tool selection discipline — Knows which tool solves which problem; doesn't default to one tool
• Claude (Chat) — Daily reasoning, requirements translation, problem-solving conversations
• Claude (Code) — Application architecture, implementation from requirements, iterative building
• Claude (Design) — UX reasoning, interaction design, visual direction
• ChatGPT — Thinking partner for regulated work, early-stage ideation, conversational problem-solving, personal artwork exploration
• Microsoft Copilot — Enterprise infrastructure mining, learning from existing team processes, MS-ecosystem implementation

Agentic & Automation Skills

• LinkedIn sweep automation — 16 searches, three-gate triage, GitHub JSON sync; runs independently with human review
• Board sweep automation — Visual ATS page reading, multi-platform support, unified aggregation
• Infrastructure that doesn't depend on memory — Coordination systems, documentation that outlasts the session, GitHub workflows with decision logs

Prompt Engineering & AI Reasoning

• Chain-of-thought reasoning — Breaking complex problems into step-by-step logic
• Few-shot prompting — Using examples to guide AI output
• Role prompting — Assigning personas/contexts to shape response
• Constraint-driven design — Using hard constraints (governance, format, tone) to direct output
• Iterative refinement — Multiple rounds of prompt/feedback to improve outputs
• Structured use-case documentation — Recording business problem → AI task → human role → output → benefit

Enterprise AI Application

• Copilot for research mining — Finding existing solutions and patterns within enterprise Teams/SharePoint/product documentation
• Copilot for MS-environment implementation — Adapting research into working solutions using Microsoft tools
• ChatGPT for workflow design — Building AI-assisted processes in regulated industries (90% time-to-draft reduction, 360training)
• Power BI + Copilot integration — Aggregating data and creating executive visibility dashboards
• Stakeholder communication with AI — Translating complex work into clear messaging for different audiences

Strategic & Process Skills with AI

• Problem diagnosis before tool selection — Identifies what's broken, then chooses the right AI (or no AI)
• Knows when AI doesn't help — Recognizes problems that need process redesign, not automation
• Governance & compliance thinking — Designs AI workflows within regulatory constraints (regulated copy, healthcare, confidentiality)
• Teaching AI workflows to others — Documenting and handing off processes so they outlast your involvement
• Design for human reality — Builds applications that account for how people actually work

Systematic Learning & Credentials

• Four LinkedIn Learning Certificates — Azure AI Essentials, AI for Project Managers, Project Management with Copilot, Microsoft 365 Copilot Quick Tips
• Nine-week structured learning plan (in progress) — Prompt fundamentals, vibe coding, agentic automation, Power BI
• Longer-term learning roadmap — Harvard CS50x, Google ML Crash Course, 3Blue1Brown, Hugging Face, MIT Missing Semester

Production Applications Built

• Job Search Command Center — Dashboard treating applications as a managed program; 11 versions, 5+ months of daily iteration, 33% Tier 1 signal rate
• Journal App — Flask, AES-256-GCM encryption, SQLite, PyInstaller, Apple Health integration; privacy-first, local-only
• Learning Platform — Hypothesis-driven approach to documenting AI skill-building; Phase 1 launched, active daily use
• Family Photo-sharing App — Loveable → OpenCode pivot; now in beta with Claude Code. Earlier versions served as demos while defining the problem space.

All maintained actively, not portfolio pieces.

How I think about this

I don't lead with the tool. I lead with the problem.

What's actually broken? What's the meaningful goal? What would a solution look like? Then: which tool accelerates this part of my work?

I study this systematically. I completed four LinkedIn Learning certificates: Azure AI Essentials, AI for Project Managers, Project Management with Copilot, Microsoft 365 Copilot Quick Tips. I'm working through a nine-week structured learning plan -- Weeks 1-2 on prompt fundamentals, Weeks 3-5 on vibe coding and building tools, Weeks 6-9 on agentic automation systems. Fridays in parallel I'm building independent Power BI analyses.

I document every AI use case: the business problem, what I used AI for, my editing role, the output, the time or quality benefit. That documentation lets me know which problems AI actually solved and which ones made things harder.

But the foundation isn't the tools. The foundation is knowing what problem you're solving. After twenty years of that practice, AI fits into the work the same way every other tool does -- it accelerates parts that don't require judgment, so judgment can do its job.