Use Case Intelligence RSS Search use cases Search
Google Cloud Retail France

Carrefour

Enterprise Knowledge Search · IT Operations Automation

Maximum cost-efficiency with scale-to-zero: Hosting on Cloud Run ensures the infrastructure completely shuts down at night and on weekends, reducing operational costs to the absolute minimum

Read the original story Explore all use cases
Carrefour transformed internal support by deploying a serverless, RAG-based AI assistant in Google Chat, reducing resolution times from hours to under two minutes while resolving 75% of queries.
How do you scale platform support for thousands of internal data users without burning out your engineering team?
Maximum cost-efficiency with scale-to-zero: Hosting on Cloud Run ensures the infrastructure completely shuts down at night and on weekends, reducing operational costs to the absolute minimum
This serverless model is particularly efficient because it scales completely to zero during inactive hours, eliminating idle compute costs.

Reported impact

Business outcome

With a multi-format network of more than 15,000 stores in more than 40 countries, the Carrefour Group is one of the world leaders in food commerce.

  • Cost Reduction
  • Productivity Improvement
  • Speed to Market

Solution

Integration effort is the true heart of the project: The most complex development did not focus on prompt engineering, but on creating a robust, asynchronous middleware (FastAPI and Cloud Run ) capable of preventing Google Chat API timeouts State management with Agent Platform sessions: Integrating this component is essential to preserve conversational context during complex, multi-turn SQL troubleshooting queries Total observability through BQAA (BigQuery Agent Analytics) : Streaming agent events to BigQuery allows the team to monitor performance, analyze user behavior, and control API costs in real-time Maximum cost-efficiency with scale-to-zero: Hosting on Cloud Run ensures the infrastructure completely shuts down at night and on weekends, reducing operational costs to the absolute minimum

Solution architecture

6 components · 4 layers
  1. Compute
    • Cloud Run Runs serverless application services and APIs with automatic scaling for bursty workloads.
  2. Storage
    • BigQuery Serves as the governed analytics and storage layer for high-volume operational data, features, and reporting.
  3. Serving
    • BigQuery Managed Context Protocol (MCP) server Supports a core part of the deployed AI and data workflow described in the customer story.
    • BigQuery Agent Analytics (BQAA) Supports a core part of the deployed AI and data workflow described in the customer story.
  4. Orchestration
    • Agent Development Kit (ADK) Coordinates multi-agent workflows, tool calling, and governed automation across the application.
    • Agent Platform Coordinates multi-agent workflows, tool calling, and governed automation across the application.

Architecture clues

  • Carrefour transformed internal support by deploying a serverless, RAG-based AI assistant in Google Chat, reducing resolution times from hours to under two minutes while resolving 75% of queries.
  • Integration effort is the true heart of the project: The most complex development did not focus on prompt engineering, but on creating a robust, asynchronous middleware (FastAPI and Cloud Run ) capable of preventing Google Chat API timeouts
  • Scaling platform support for thousands of internal data users without burning out your engineering team is a massive challenge.
  • To ensure rapid deployment, high observability, and strict cost control, the team leveraged a vertically integrated, serverless Google Cloud stack.

Evidence from the source

State management with Agent Platform sessions: Integrating this component is essential to preserve conversational context during complex, multi-turn SQL troubleshooting queries
The analytical brain is powered by Gemini 2.5 Pro, selected for its advanced reasoning capabilities and European regional availability. For the retrieval pipeline, Agent Platform Search (Discovery Engine) indexes interna
The entire setup is hosted on Cloud Run, which runs both the middleware and the ADK code. This serverless model is particularly efficient because it scales completely to zero during inactive hours, eliminating idle compu
The middleware receives the HTTP POST webhook from the Google Chat API and immediately returns a 200 OK status to the platform. Simultaneously, it triggers a background worker task using FastAPI's BackgroundTasks utility
To ensure rapid deployment, high observability, and strict cost control, the team leveraged a vertically integrated, serverless Google Cloud stack. At the center of orchestration is the Google Cloud ADK, which coordinate
Transparency Evidence & methodology 76/100 reference value · 77% confidence · 1 source

Review how this reference was scored and trace each normalized claim back to its documented source.

01

Scoring methodology

How scores are calculated

Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.

Customer reference value76/100
  • Problem clarity20/20
    Measures whether the customer's operational problem is specific enough to compare.
  • Implementation depth15/25
    Rewards named components, data flow, integrations, and architecture detail.
  • Outcome specificity16/20
    Measures whether the reported benefit is concrete rather than a generic claim.
  • Quantified impact15/25
    1 metric describe measurable impact.
  • Deployment scale10/10
    Rewards evidence of a live or scaled deployment and meaningful integration scope.
Evidence confidence77%
  • Source authority20/25
    The main evidence is an official vendor-published customer story.
  • Claim traceability21/25
    10 of 12 key fields link to a named source.
  • Direct evidence20/20
    5 direct source quotes retained.
  • Date provenance0/10
    The publication date has no documented provenance.
  • Source transparency8/10
    1 source reference document publisher, URL, authority, retrieval, and supported fields.
  • Implementation specificity8/10
    Rewards source-backed components, data flow, quantified impact, and a specific outcome.
02

Traceability

Sources & provenance

Field coverage shows which facts each source supports. Dates are shown only when documented by the source.

  1. Google Cloud

    Carrefour customer story

    First-party
    https://cloud.google.com/customers/carrefour-tech
    Source type
    Vendor-published customer story
    Supports
    • Customer
    • Business problem
    • Solution
    • Products
    • Solution components
    • Data flow
    • Integrations
    • technical area
    • use case category
    • Business outcome
    • outcome category
    • Metrics
    • Architecture
    • evidence quotes

Date evidence

  • Sitemap last modified Aug 20, 2026 · Google Cloud · Publisher metadata Publisher-reported modification timestamp; not a publication date.
    Open date source
  • Unverified extracted date Aug 20, 2026 · Google Cloud · Unknown Legacy extracted date without source-page confirmation.
    Open date source
Not separately sourced Industry, Region, Company size. Treat these normalized attributes as lower-confidence until a dedicated source is added.

Related

Similar use cases

See all like this
Google Cloud Archived · Aug 19, 2026 Wayfair Retail India Wayfair needed to reduce developer context switching and long CI failure recovery times by giving engineers fast, contextual explanations and fixes inside the tools they already use. 58% reduction in Mean Time To Recovery (MTTR), from 26 minutes to 11 Enterprise Knowledge SearchIT Operations Automation Buildkite · Gemini · Gemini platform Google Cloud Modified · Aug 6, 2026 Trade Me Retail New Zealand Trade Me wanted to turn a marketplace and engineering organization into an AI-first ecosystem, which meant replacing physical data centers, accelerating incident diagnosis, and letting employees query marketplace trends… Achieved 4 9s availability for critical APIs through Enterprise Knowledge SearchIT Operations Automation Agent Development Kit · AlloyDB · BigQuery Google Cloud Modified · Jun 26, 2026 Kogan.com Retail Australia Kogan.com needed to eliminate brittle export-based BI pipelines, shorten slow answer cycles for business users, and expose analytics conversationally without losing semantic consistency or data-level security. 150 non-technical team members Enterprise Knowledge SearchIT Operations Automation BigQuery · Looker · Model Context Protocol (MCP) fleet server