RSM to the Power of AI
My work builds the internal AI capability that helps make all three possible.
My Role
I build the internal AI capability that makes RSM possible: empowering people to work with AI, helping them build their own tools, creating trusted guidance and governance, standardizing what works, and scaling it across eBay.
From Watching AI to Building With It.
People Team-wide sessions to share what works and help others reuse it.
- ✓ Real examples
- ✓ Success stories
- ✓ Practical demos
- ✓ Lessons learned
Hands-on workshops focused on building real agents step by step.
- 01 Identify a useful workflow
- 02 Create the first version
- 03 Add trusted sources
- 04 Find errors and edge cases
- 05 Refine and improve
Full-day sessions customized around the work of different departments.
Learning Model
What teams leave with
- Working prototypes
- Reusable prompts
- Agent-building skills
- Practical AI workflows
- Confidence to keep building
Four Efforts. One Goal.
Move from individual AI experimentation to scalable AI capability.
Guidance + Governance + Platforms + Adoption
Employees were navigating a growing list of AI tools, scattered resources, rapid policy changes, and unclear data classifications with no single place to start. Good intent, no clear path.
Tool + model selection guides employees to the right path
Clear → Guide · Complex → Expert · Not approved → Alternatives
Employees start by describing what they're using and what for — not by reading a policy document.
Every path ends with guidance, an approved alternative, or a direct connection to the right expert.
Guidance links back to the maintained policy or tool standard — not a static document that goes stale.
Complex or gray-area questions route directly to the right team: Privacy, Responsible AI, or GCX.
Roadmap
- Tool guidance
- Data classification
- Source links
- Approved alternatives
- Expert handoffs
- Tool + Model + Data + Use Case
- Improved coverage
- Maintained sources
- Evolving guidance
- Ownership
- Lifecycle reviews
- Governance
- Reusable standards
- Continuous improvement
- Pilot feedback described the experience as useful, innovative, and worth developing further
- Partners emphasized the value of approved alternatives over a simple "not allowed" dead end
- Demand for broader coverage confirmed in early testing — validating the phased roadmap approach
Find the opportunity before choosing the AI. Partnering to understand where teams have real workflow gaps, what AI can realistically support, and where to prioritize first.
Build guardrails into the work, not onto the end. Connecting responsible-use expectations with how internal AI agents are actually designed, tested, and maintained.
Reuse the learning. Improve the experience. Partner on the eBay product catalog, share checker logic with HubGPT standards, and explore AI-assisted component development with engineering.
Five Principles I Won't Compromise On.
These shape every AI effort on this page, from the workshops to the governance work to the Tools Checker. They're not rules I follow. They're decisions I make repeatedly.
Every effort on this page started with a real employee problem: too many tools, no clear next step, scattered guidance. The AI came after the need was clear. Always.
The workshops weren't designed for passive learning. Every participant leaves with something they built themselves. The goal is teams that keep going after I leave the room.
The AI Tools Checker's core design rule: never leave someone without a path. Clear answer? Guide them. Complex case? Connect them to an expert. Not approved? Show alternatives. The experience ends with a step, not a shrug.
Guardrails are much easier to build in than bolt on later. Every company-level effort on this page brought Privacy and Responsible AI into product decisions from the start, not at the end of the process.
The checker logic, the governance model, the agent-building patterns across the workshops: all designed to be shared. When one team figures something out, the next team shouldn't have to start from scratch.
Calls I Made That Weren't Obvious.
Leadership isn't just strategy. It's the specific decisions you push through, push back on, and own the consequences of.
The default path: build the AI experience, then route it through approval. I pushed for a different model. Privacy and Responsible AI became design partners from day one on every company-level effort. Slower at first. No major rework after.
The easy path was awareness sessions: demos, overviews, Q&A. I held the line that every workshop ends with something participants built themselves. Teams that only watch don't keep going. Teams that build do.
Early versions tried to answer every AI question at once. I cut it back to high-confidence answers only, with clear routing for everything else. A tool that tries to cover everything and gets things wrong loses trust fast. A tool that knows its limits earns it.
AI · Product · UX Certifications
Each certification informed a specific decision or workflow change at eBay.
To discuss AI transformation leadership, the frameworks I've built, or how I approach responsible AI at scale, connect on LinkedIn / Nancy-UX.