Gemini Enterprise Agent Platform & Google ADK
I help teams with real, repeatable processes build agents on Google's Gemini Enterprise Agent Platform — with the code-first Agent Development Kit (ADK) when you want to build in code. For teams already on Google Cloud, it's typically my default. If Claude fits your case better, I'll tell you honestly.
A short call to see whether agents on Google Cloud — or Claude — make sense for your case. No sales pitch.
What is the Gemini Enterprise Agent Platform?
The Gemini Enterprise Agent Platform is Google's managed stack for building and running production AI agents — the evolution of Vertex AI — pairing Agent Runtime, Memory Bank, Agent Identity and Agent Gateway with the code-first Agent Development Kit (ADK). ADK is the part you reach for when you want to build agents in code rather than configuration, with open standards like MCP and A2A built in.
For teams already on Google Cloud it's usually my default, because identity, governance and ecosystem fit come for free. If Claude fits your case better, I'll tell you honestly — these are equivalent platforms with real tradeoffs, not a winner and a loser.
Why the Gemini Enterprise Agent Platform?
A platform, not just a framework
The Gemini Enterprise Agent Platform is the evolution of what used to be called Vertex AI: Agent Runtime, Memory Bank, Agent Identity and Agent Gateway, plus the code-first Agent Development Kit (ADK) for builders. In my experience it's a strong default for teams already on Google Cloud — here's why:
Identity & Governance
Agent Identity and Agent Gateway give you the access controls and auditability that enterprise IT typically expects from day one.
GCP Ecosystem Fit
If your data and services already live on Google Cloud, the platform tends to slot in with far less integration glue than bolting an agent stack on from outside.
Code-First with ADK
When configuration isn't enough, ADK lets you build agents in code — versioned, testable, and owned by your team rather than locked in a console.
Open Protocols
MCP for tools and A2A for agent-to-agent communication mean you're building on open standards, not a closed island.
None of this makes Google the only answer. Claude (Agent SDK, Code) is an equally legitimate choice with its own tradeoffs — I'll help you weigh both honestly.
How I Can Help
Discovery & Strategy
- Current architecture review and agent fit
- Gemini Enterprise Agent Platform vs Claude — an honest call
- Where ADK (code-first) makes sense vs configuration
- Integration roadmap across your Google Cloud stack
- A rough cost and effort outline (not a fixed quote)
Agent Development on ADK
- Custom agents using ADK's Runner architecture
- Multi-agent orchestration with the A2A protocol
- Tool integrations via MCP
- Memory and context management with Memory Bank
- Prompt and evaluation work, versioned in code
Platform Deployment
- Agent Runtime deployment on the Gemini Enterprise Agent Platform
- Agent Identity and Agent Gateway for access control
- Security and compliance configuration
- Observability, monitoring and cost visibility
- Team enablement and knowledge transfer
Ongoing Advisory
- Fractional CTO support as your agents evolve
- Architecture reviews and second opinions
- Help keeping options open across Google and Claude
- Coaching your team to own what we built together
Most engagements start small with a Discovery, so we can confirm agents are the right tool before committing to a build.
What the Platform Gives You
Core ADK Architecture
Runner-Based Execution Model
- • Event-driven architecture for scalable agent operations
- • Asynchronous processing with built-in error handling
- • State management across distributed agent networks
- • Native Google Cloud integration for enterprise scale
A2A Protocol Mastery
- • Inter-agent communication standards
- • Message routing and orchestration
- • Consensus mechanisms for multi-agent decisions
- • Fault tolerance and recovery patterns
Working with Gemini
- •Model Selection: Flash or Pro depending on the use case
- •Prompt & Eval Work: patterns kept versioned in code
- •Cost Awareness: keeping token usage sensible
- •Model Flexibility: ADK isn't locked to a single model
Google Ecosystem
- • Agent Runtime for serving and scaling agents
- • Cloud Functions for serverless execution
- • BigQuery for data processing
- • Cloud Storage for knowledge bases
- • Workspace APIs for productivity automation
Enterprise Systems
- • ServiceNow and similar ITSM platforms
- • SAP and Oracle ERP systems
- • Microsoft 365 via Graph API
- • Custom enterprise applications via MCP tools
- • Your own APIs and internal services
Where Agents Tend to Earn Their Keep
Illustrative patterns, not client benchmarks — outcomes depend heavily on your data and process
Back-Office Document Processing
Challenge
High volumes of documents needing classification and data extraction
Solution
ADK agents with Gemini handling intake, extraction and routing
Results
- Less manual triage
- Consistent extraction
- Multi-format support
Customer Support Assist
Challenge
Repetitive inbound questions spread across several channels
Solution
Agents that draft answers and route the hard cases to humans
Results
- Faster first responses
- Humans kept in the loop
- Audit trail of actions
Operational Monitoring
Challenge
Manual checks across suppliers, systems or logistics partners
Solution
Agents that watch signals and surface anomalies for review
Results
- Earlier issue detection
- Fewer missed alerts
- Clearer visibility
Internal Knowledge Agents
Challenge
Teams hunting through wikis, tickets and docs for answers
Solution
Grounded agents over your own content via MCP-connected tools
Results
- Answers with sources
- Less context-switching
- Stays within your data
Who You'd Be Working With
Background
- •Fractional CTO: based in Brescia, Italy; GDG Brescia organizer
- •Google Partner (through MaGa Srl):Google Cloud Select Co-sell Partner and Google Workspace Select Co-sell & Services Partner
- •Experience: over 20 years, 250+ projects across web, mobile and cloud
- •Two platforms, no favourites: Google (Gemini Enterprise Agent Platform + ADK) and Claude (Agent SDK, Code)
ADK-Specific Skills
- •Runner architecture patterns
- •A2A Protocol implementation
- •MCP tool integrations
- •Agent Runtime deployment
- •Google Cloud architecture
Development Approach
How a Build Typically Runs
A rough shape, not a fixed plan — Build & Deploy is usually 8–16 weeks depending on scope
Discovery & Architecture
- •Business process analysis
- •Agent fit and Google vs Claude call
- •Architecture design with Google best practices
- •Integration points identification
- •Success criteria definition
Prototype Development
- •Core agent implementation with ADK
- •A2A protocol setup
- •Gemini model configuration
- •Initial MCP tool integrations
- •Evaluation harness
Production Preparation
- •Security hardening
- •Scalability testing
- •Monitoring setup
- •Documentation creation
- •Team training
Deployment & Advisory
- •Production rollout on Agent Runtime
- •Monitoring and cost visibility
- •Iteration with your team
- •Ongoing advisory if useful
How I Work
Discovery & Strategy
2–4 weeks
- Architecture and agent-fit review
- Google vs Claude — an honest recommendation
- Where ADK (code-first) helps
- Integration roadmap
- A rough effort and cost outline
Build & Deploy
8–16 weeks
- Agents built with ADK
- MCP tool and A2A integrations
- Deployment on Agent Runtime
- Monitoring and cost visibility
- Team enablement along the way
Ongoing Advisory
3–12 months
- Fractional CTO support
- Architecture reviews and second opinions
- Help keeping options open across platforms
- Coaching so your team owns the system
Concrete pricing comes after a short Discovery, once I understand your case — I'd rather scope it honestly than quote a number blind.
Agents vs Traditional Automation
A rough, generalized comparison — for some jobs classic RPA is still the better tool
| Feature | Google ADK | Traditional RPA | Other AI Frameworks |
|---|---|---|---|
| Setup Time | 2-4 weeks | 3-6 months | 6-12 weeks |
| Maintenance | Adapts to change | Constant updates | Regular retraining |
| Scalability | Scales with Cloud | Limited | Depends on infrastructure |
| Intelligence | Gemini-powered | Rule-based | Model-dependent |
| Cost Model | Usage-based | License + Implementation | Varies widely |
| Integration | Native Google + APIs | Complex connectors | Custom development |
Questions, Answered Honestly
The Gemini Enterprise Agent Platform is the evolution of what used to be called Vertex AI: Agent Runtime, Memory Bank, Agent Identity and Agent Gateway, plus the code-first Agent Development Kit (ADK). ADK is the part you reach for when you want to build agents in code rather than configuration.
How I'd Approach a Build
A worked example to show how I think — illustrative, not a named client case study
A Typical Starting Point
Say a team is drowning in manual document handling — intake, classification and data entry that's slow and error-prone. The first question I ask isn't “which model?” but “is an agent actually the right tool here, or would simpler automation do?”
How I'd Build It
- • A small set of focused ADK agents, not one monolith
- • Tools connected via MCP to your existing systems
- • Gemini for extraction and classification
- • An evaluation harness so we can measure quality before rollout
How I Work
A few principles I hold to on every engagement
Honest fit first
If agents — or AI at all — don't make sense for your case, I'll say so. I'd rather lose a build than ship something that doesn't earn its keep.
Two platforms, no favourites
Google (Gemini Enterprise Agent Platform + ADK) and Claude (Agent SDK, Code) are both real options. I pick based on your stack, not on a vendor I'm partial to.
Your team owns it
I build in code, version it, and coach your team so you can extend and operate the system without me.
Credentials
Let's See If Agents Fit Your Case
A short Discovery call is the honest place to start: we look at your stack, weigh Google against Claude, and decide together whether a build is worth it. No pressure, no pitch.
I take on a limited number of builds at a time
Quick Booking
Book directly on calendar✉️ Email
matteo@gazzurelli.com🌍 Location
Brescia, Italy (serving global clients)
Agents on Google's Gemini Enterprise Agent Platform — and Claude when it fits
I'm a Fractional CTO who helps teams design, build and deploy agents that survive contact with production. On Google, that usually means the Gemini Enterprise Agent Platform — the evolution of what used to be called Vertex AI — with the code-first Agent Development Kit (ADK) when you want to build in code rather than configuration.
For teams already on Google Cloud it's often my default, thanks to identity, governance and ecosystem fit. But Claude (Agent SDK, Code) is an equally legitimate choice, and I'll recommend whichever genuinely fits your case — or tell you honestly if agents aren't the right tool at all.
Book a Discovery Call
Select a time that works for you
Last updated: May 2026