What challenges did Shopify face before AI implementation?

Serving millions of merchants with an AI assistant that maintains conversation quality, handles complex multi-tool calls, and provides expert-level guidance. Need for scalable and highly available cloud infrastructure to enable rapid feature iteration and update deployment. Supporting personalized insights for merchants to improve sales performance and decision making, including for new entrepreneurs.

What was the impact of the AI solution?

Faster development and iteration cycles leading to rapid feature deployment. High uptime and resiliency supporting millions of merchants globally. Improved merchant decision-making through personalized AI insights. New entrepreneurs reach first sale significantly faster, sometimes in days instead of weeks.

Source

GCPScaled productionEvidence: Medium55/100

Shopify AI-enabled Commerce Assistant

Use case typeConversational supportUpdated Jun 13, 2026

Shopify developed Sidekick, an AI-powered commerce assistant to help millions of merchants with personalized, expert guidance for business decisions and operations. Sidekick leverages Anthropic's Claude large language models hosted on Google Cloud Vertex AI to provide natural language conversational AI that can execute multiple tool calls for complex queries. Google Cloud infrastructure components such as Bigtable, BigQuery, Compute Engine, and Google Kubernetes Engine support low latency, high availability, and rapid deployment and iteration of new AI features for Sidekick. This AI assistant enables merchants to quickly get actionable insights, facilitating faster store setup, analytics, and decision-making, speeding time to first sale for new entrepreneurs. Shopify utilizes Model Garden on Vertex AI to easily access and deploy various Claude and Gemini models internally for efficiency and innovation.

Organization: Shopify
Industry: Retail
Location: Canada
Published: May 2026

Reported outcomes

Strategic outcomes

New product / capabilityLaunched an AI commerce assistantNew product / capabilityEnabled complex multi-tool conversationsSpeed & agilityFaster feature deployment and iterationBetter decisions & insightImproved merchant decision-making

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Shopify
Provider: GCP
Maturity: Scaled Production
Linked source: Google Cloud Customer Stories

Utilized Google Cloud Bigtable, BigQuery, Compute Engine, and Kubernetes Engine to deploy and operate the platform at scale with low latency and high uptime

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 2 of 2

1AI commerce assistant
2Conversational AI for retail

Serving millions of merchants with an AI assistant that maintains conversation quality, handles complex multi-tool calls, and provides expert-level guidance.
Need for scalable and highly available cloud infrastructure to enable rapid feature iteration and update deployment.
Supporting personalized insights for merchants to improve sales performance and decision making, including for new entrepreneurs.

Built Sidekick using Anthropic Claude LLMs hosted on Google Cloud Vertex AI to leverage advanced reasoning and tool-calling capabilities.
Utilized Google Cloud Bigtable, BigQuery, Compute Engine, and Kubernetes Engine to deploy and operate the platform at scale with low latency and high uptime.
Implemented Model Garden on Vertex AI to facilitate easy access to multiple AI models for internal tool building and faster development cycles.
Maintained conversation quality and performance even with complex multi-turn conversations involving multiple tool invocations.

Technologies

Claude on Vertex AI Vertex AI Bigtable BigQuery Compute Engine Google Kubernetes Engine

Faster development and iteration cycles leading to rapid feature deployment.
High uptime and resiliency supporting millions of merchants globally.
Improved merchant decision-making through personalized AI insights.
New entrepreneurs reach first sale significantly faster, sometimes in days instead of weeks.

Sources & evidence1

Evidence: Medium55/100Evidence strength