On Establishes Google Cloud as Enterprise AI Backbone, Beginning with Agent-Led Cloud Migration

Premium sportswear brand uses a multi-agent AI architecture to migrate 24 critical services to Google Cloud in weeks, establishing the foundation to modernize data platforms and deploy Gemini

Google Cloud and premium sportswear brand On today announced that On used AI agents to migrate 24 core services to Google Cloud, cutting migration time from three months to just two weeks per service. A small team of On’s internal engineers ran 15 migrations completely in-house through the new agent-led approach with under five minutes of planned downtime per service. Consolidating data, compute, and AI on Google Cloud has enabled On to scale workforce productivity and build custom internal agents using Gemini Enterprise.

As On scaled globally, the challenge was no longer simply running infrastructure efficiently, but creating a technology foundation capable of supporting the next phase of its growth. That meant reducing fragmentation between compute, data and AI, simplifying the architecture, and creating a common platform from which teams could build and deploy intelligent workflows and agents. On selected Google Cloud as that foundation.

“AI changes the economics of execution. We’re seeing that firsthand when our engineers used AI agents to deliver a major cloud migration with no additional resources and zero business disruption,” said Narek Verdian, chief technology officer, On. “The real opportunity is rethinking how technology work gets done, while ensuring human engineers remain accountable for the decisions that matter.”

On’s engineers saw productivity gains right away. Rather than relying on external systems integrators, internal engineers used agentic workflows to map complex, inconsistent deployment patterns across more than 20 code repositories in hours instead of days. By offloading mechanical tasks such as configuration scripting, data replication, and environment validation to AI agents, On protected its product engineering roadmap, eliminated the need for outside implementation support, and limited planned production cutover downtime to under five minutes per service.

“True agentic transformation requires a unified data and compute foundation,” said Karthik Narain, chief product and business officer, Google Cloud. “On proved that specialized AI agents can fundamentally shift the efficiency and velocity of complex cloud migrations, setting the stage to modernize their critical data platforms on Google Cloud. Now, with Google Cloud serving as their enterprise AI backbone, On is using Gemini Enterprise to build and deploy custom agents that drive real business impact across their workforce.”

Modernizing infrastructure through agentic AI

As On expanded globally, the premium performance brand faced a major engineering challenge: modernizing and migrating complex, deeply integrated microservices to Google Cloud without disrupting retail operations or slowing development.

To address this challenge, On’s technology team and Google Cloud deployed a new paradigm for executing technical work using AI agents. Moving beyond standard developer assistance tools, toward coordinated, domain-specific AI agents capable of handling complex migration tasks end-to-end, On was able to work from one single agentic platform, leveraging secure and powerful models.

Bounded execution and human oversight

The multi-agent system divided the migration into specialized, automated steps, including codebase analysis, infrastructure configuration generation, translation pipeline creation, and testing validation. Engineers reviewed generated code, ran infrastructure changes, and authorized production cutovers. Agents were given defined tasks with clear checks, while consequential changes remained subject to human approval.

By shifting execution speed and capability to agentic AI, On unlocked substantial resource efficiency. For example, an external vendor initially submitted a $500,000 quote to migrate just a fraction of the microservices using traditional manual methods, On’s internal engineering team was able to complete the broader project far faster without needing outside implementation resources.

The success of this migration showcases that in the era of agentic AI, constraints have shifted and the economics of this change pays off. This opens up the possibility to execute other complex workflows following a reusable blueprint, rethinking the work from the ground up, instead of automating a process that already exists.

On’s AI ambitions go beyond its technology platform. The company rolled out Gemini Enterprise to every employee, empowering teams to build custom, no-code agents for executive reporting, market research, scheduling, onboarding, and more. It has also been a strategic partner to Google Cloud, providing feedback as a trusted tester  of the new  Gemini agent, Google Cloud’s single universal agent for work.