Use Case Intelligence RSS Search use cases Search
Google Cloud Logistics Germany

Delivery Hero

Application Refactoring · IT Operations Automation

Delivery Hero needed a standardized AI engineering platform that could support autonomous software delivery, let teams use the strongest coding model for each task, and still preserve centralized governance, review, and security across a global engineering organization.

Read the original story Explore all use cases

Reported impact

Business outcome

Delivery Hero established a governed multi-model engineering platform that increases developer autonomy, supports autonomous feature work, and keeps review and security embedded in the software-delivery flow.

  • Productivity Improvement
  • Risk Reduction
  • Speed to Market

Solution

Delivery Hero built its agentic engineering platform on Google Cloud using Gemini Enterprise Agent Platform as the shared control layer for multiple coding and review models. The company created Herogen, an autonomous software delivery agent, on top of Claude and Google Cloud infrastructure so engineers could move beyond autocomplete toward guided feature completion. Agent Platform's open design let Delivery Hero place LiteLLM in front of the environment, giving developers IDE-agnostic access to models such as Claude Opus and Gemini while the central team maintained one governed AI foundation. Claude handles most coding tasks, while Gemini participates in a council-of-agents review flow that checks code quality and security before changes reach production. Delivery Hero also uses Gemini Code Assist and Gemini CLI for day-to-day development work, turning Google Cloud into the backbone for both autonomous delivery and governed multi-model software engineering.

Data flow

Engineer requests or autonomous Herogen tasks enter through LiteLLM into Gemini Enterprise Agent Platform. Claude generates or advances the code, Gemini participates in review and security checks, and approved outputs return to the engineer's IDE or delivery workflow.

Solution architecture

6 components · 3 layers
  1. Compute
    • Claude Serves as the primary coding model for Herogen and day-to-day software-development tasks.
    • Gemini Reviews generated code and performs additional quality and security checks in a multi-agent workflow.
  2. Serving
    • LiteLLM Acts as an LLM proxy so engineers can access multiple models from their preferred IDEs.
    • Gemini Code Assist Supports general software-development tasks when teams need a cost-effective coding assistant.
    • Gemini CLI Extends Google Cloud-assisted development into command-line engineering workflows.
  3. Orchestration
    • Gemini Enterprise Agent Platform Provides the enterprise AI control plane that standardizes model access and powers autonomous engineering workflows.

Architecture clues

  • Claude is the primary coding model while Gemini participates in a council of agents
  • Engineers can access models from their preferred IDEs
  • Herogen runs as an autonomous software delivery agent
  • LiteLLM is used as an LLM proxy in front of Agent Platform

Evidence from the source

Delivery Hero chose Google Cloud’s Agent Platform.
Delivery Hero uses Claude as the primary model for coding tasks, while Gemini serves within a 'council of agents' to review the code and conduct security checks.
Delivery Hero uses Gemini Code Assist and Gemini CLI
The platform’s open nature allowed the team to implement LiteLLM as an LLM proxy.
The power of Herogen lies in the tandem use of Claude and Google Cloud models.
Transparency Evidence & methodology 70/100 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 value70/100
  • Problem clarity20/20
    Measures whether the customer's operational problem is specific enough to compare.
  • Implementation depth20/25
    Rewards named components, data flow, integrations, and architecture detail.
  • Outcome specificity20/20
    Measures whether the reported benefit is concrete rather than a generic claim.
  • Quantified impact0/25
    0 metrics 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

    Delivery Hero customer story

    First-party
    https://cloud.google.com/customers/deliveryhero
    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

  • First archived Sep 26, 2026 · Internet Archive · Independent First Wayback Machine capture; confirms the page existed by this date, not when it was published.
    Open date source
  • Sitemap last modified Jun 29, 2026 · Google Cloud · Publisher metadata Publisher-reported modification timestamp; not a publication date.
    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 Modified · Oct 9, 2026 Porter Logistics India Porter (Smartshift Logistics Private Limited), a leading logistics platform based in India, orchestrates on-demand mini-trucks, tempos, and two-wheelers across 50+ major cities. When order volumes doubled, Porter’s… Achieved 30% zero-touch driver verification automation on Gemini platform Application RefactoringDocument Processing BigQuery · Cloud Composer · Cloud Pub/Sub Google Cloud Archived · Aug 29, 2026 PayPal Financial Services United States After unifying its core data estate, PayPal needed an AI platform that could power global commerce use cases, let developers choose the best model for each workload, and elastically handle training and inference without… 25,000 pipelines every day Application RefactoringIT Operations Automation BigQuery · Cloud GPUs · Gemini Enterprise Agent Platform Google Cloud Archived · Aug 29, 2026 WeSpeak Software / SaaS Argentina Building that product required infrastructure that would never go down. Early on, WeSpeak ran on a virtual private server that could fail without warning, taking the entire platform offline. For a hotel waiting on a… 70% of hotel guest inquiries automated end to end Application RefactoringIT Operations Automation Cloud Run · Gemini · Gemini Flash