August 2017. Gap Inc.’s Director of Cloud Infrastructure resigned.
Four teams — IaaS, Private Cloud, PaaS, and Tooling — were left leaderless, demoralized, and behind schedule on a critical migration to private cloud infrastructure. Two engineers had already decided to leave before I arrived. Both said they were sad about it — they would have stayed if the timing had been different.
13 weeks until Black Friday.
For a $15B retailer, PEAK weekend carried outsized operational and revenue risk — the kind of failure that makes headlines and moves annual numbers. Previous PEAK weekends had already produced outages and lost sales. The new private cloud infrastructure needed to be ready to absorb overflow ecommerce traffic. It wasn’t. The team building it had stopped functioning.
The technical problem was solvable. The human problem wasn’t — yet.
The traditional VMware/DevOps teams in Pleasanton viewed private cloud as an existential threat to their jobs. Gap was already decommissioning co-hosted metal infrastructure. At the individual engineer level, relationships were fine. At the leadership level there was no communication — only distance, physical and organizational. My teams were at SF HQ. The traditional infrastructure leadership sat in Pleasanton.
Four broken teams. Hostile peer organizations. Two engineers who had already decided to leave. Thirteen weeks to the highest-stakes weekend of the year, with the CTO and CIO planning to watch live.
I started with the engineers, not the roadmap. Weekly 1:1s with every team member — understanding what motivated them, what blocked them, what they wanted to build. I gave people meaningful work and visible paths forward. Attrition stopped.
The traditional infrastructure directors in Pleasanton didn’t need to be negotiated with — they needed to be understood. They were afraid their teams were being made obsolete, and they weren’t entirely wrong. I named that directly and found genuine common ground. Coordination replaced silence.
These were experienced professionals who were passionate about their work. Once the obstacles were cleared — the morale problems, the cross-org hostility, the leadership vacuum — they didn’t need driving. They needed to be left alone to do what they were good at. I got out of their way. They got ready.
PEAK weekend 2017: zero downtime. Zero revenue loss.
The private cloud infrastructure successfully handled overflow ecommerce traffic for the entire event. The CTO and CIO watched live dashboards in the Pleasanton war room.
The traditional infrastructure teams that had initially viewed us as a threat began treating us as trusted partners. That shift unlocked the coordination the migration had been missing.
The following year, the Director managing a fifth abandoned team sought me out and asked me to turn that one around too. When he left for AWS just before PEAK, I stepped in as interim Director for the event.