Scaling capability beyond the core team
AI design demand outpaced supply. Teams wanted to move fast, but lacked the foundational fluency to do so responsibly. The central AI team risked becoming a bottleneck.
If this failed: We would either slow everyone down through gatekeeping, or lose control of quality and consistency as teams shipped AI without shared understanding.
Capability didn't scale with demand. If we didn't teach teams how to make good AI decisions, every decision would route through us—creating dependency instead of autonomy.
If we could empower teams with the right knowledge and frameworks, they could own AI decisions confidently, and we could focus on higher-leverage platform-level work.
The core strategy was to scale judgment, not just knowledge. Teams needed more than definitions—they needed decision frameworks, pattern recognition, and confidence to apply responsible AI principles in context.
Strategic choices:
Tradeoffs I made:
Alternatives I rejected:
Multi-tiered enablement program:
Creating and documenting standards while the products were still evolving. Staying ahead was rarely possible, so I built feedback loops with partner teams and treated documentation like a product backlog—shipping iteratively based on what teams needed most urgently.
I recruited Ambassadors from teams already shipping AI successfully and ran biweekly roundtables where they could review in-flight work, discuss new patterns, and influence platform decisions through early feedback. This gave them early access to guidance and elevated their AI fluency, while helping us understand how patterns needed to adapt across different team contexts.
Over time, they became the first line of support in our community channels—answering questions, sharing examples, and modeling best practices. This distributed expertise meant my central team could focus on platform-level problems instead of repeated 1:1 pattern coaching.
Enablement only works when it's aligned with how teams ship. I partnered across Product and Engineering to tie the curriculum to real delivery moments: design reviews, pattern adoption, and governance checkpoints.
Product leadership was initially skeptical—they saw enablement as slowing down innovation in the rush to ship AI products. I reframed it: every repeated escalation costs 2-3 hours of senior design time. Investing 100 hours in shared fluency would save 300+ hours in redundant 1:1 coaching. That math changed the conversation. Four months later, Product launched their own enablement program using the same model.
Risk: time and resourcing. The central team was already shipping core AI platform experiences—adding enablement on top risked burnout and delivery delays.
Response: I designed enablement to be incremental, modular, and community-powered. We shipped smaller improvements continuously, prioritized the highest-leverage standards first, and relied on tight partner feedback loops to validate what to codify next.
Pushback: Early on, teams wanted to move fast with bespoke AI concepts that didn't fit platform patterns. I used research to separate "platform-wide patterns" from "approved custom solutions" and created a governance path that kept innovation moving without fragmenting the platform. Teams could still differentiate—they just needed to follow responsible AI standards and document their rationale.
Before: Teams depended on the AI design group for basic pattern decisions. Every question funneled to us. The support channel was one-way: we answered, they listened.
After: The community started answering itself. Ambassadors and close partners shared examples, corrected misconceptions, and helped teams apply patterns to new use cases with confidence. Our support channel became a living knowledge base where expertise was distributed, not centralized.
"Great communication, meeting facilitation, visibility into work-in-progress. High value, high return on time investment."
"Michelle has been my lifeline."
Reduced bottlenecks on the central AI team and improved first-pass quality in design reviews by 40% as teams gained shared standards, vocabulary, and reusable components. The program also drove a 30% confidence increase among design managers in their ability to evaluate and approve AI designs—ensuring leadership could guide their teams with the same rigor and shared language. Most importantly, it created an active network of AI design advocates who could teach, reinforce, and evolve the practice across the org.
I'd keep the same core strategy, but I would change three things:
I underestimated how much change management mattered. The content was necessary, but creating rituals, incentives, and leadership alignment is what made it stick.
I scale impact by turning ambiguity into systems: standards, rituals, and communities that make quality repeatable and teams more autonomous. This case demonstrates my ability to build infrastructure that outlasts individual contributions.