1.2 Platform Selection: Agent Designer, CES & The ADK Scaling Bridge

1. Key Concept Definitions: Studio Platforms & The ADK Bridge

1. Agent Designer (Low-Code Studio)

The no-code/low-code visual builder on the Gemini Enterprise Agent Platform for launching internal workforce assistants grounded on enterprise data and SaaS connectors in days.

2. Identity-Aware Agent Search

Indexes enterprise repositories (SharePoint, Google Drive, Confluence) and enforces source-system document Access Control Lists (ACLs) at query time based on the logged-in employee’s identity.

3. Pre-Built Action Connectors

Turnkey integrations (Jira, ServiceNow, Salesforce, Google Workspace) with managed per-user OAuth 2.0 consent so employees execute SaaS actions under their own permissions.

4. The Agent Designer -> ADK Bridge

Lets teams validate prompts and data stores rapidly in Agent Designer, then transition to (or delegate via A2A to) code-first ADK for callbacks, composite workflows, and Git CI/CD.

2. Interactive Platform Decision Engine & The ADK Scaling Bridge

Switch between the 3 Platform & Bridge Modes below and click the 3 stages to see how Agent Designer, Customer Engagement Suite (CES), and the Agent Designer -> ADK Bridge operate in production:

AGENT DESIGNER STAGE 1: VISUAL AUTHORING & GROUNDING

1. No-Code Visual Configuration & Identity-Aware Grounding

How It Works

Business analysts define the agent’s persona and instructions in a visual studio and attach Agent Search data stores (SharePoint, Drive, Confluence) that enforce document ACLs automatically.

Core Architecture Rule

Select Agent Designer when a non-engineering or line-of-business team needs to launch an internal employee assistant rapidly!

3. Platform Selection & Scaling Bridge Matrix

Match any team profile and business requirement to the corresponding Google Cloud platform strategy:

Architectural Dimension	Agent Designer (Low-Code Enterprise Studio)	Customer Engagement Suite (CES) (Omnichannel CCaaS)	The Agent Designer -> ADK Bridge (Low-Code to Code-First)
Primary Builder Persona	Business analysts, IT admins, and product managers (no/low-code).	Conversational UX designers and contact-center operations teams.	Cross-functional teams: Analysts prototype in Studio; Engineers scale in Python/Java ADK.
Target End-Users & Channels	Internal employees via Gemini Enterprise web app & Google Workspace.	External customers via Voice / Telephony (PSTN/SIP), Web Chat, SMS, and Mobile.	Starts as an internal pilot; scales to custom enterprise APIs or multi-agent production backends.
Orchestration & Guardrails	Natural-language instructions, Agent Search ACLs, and pre-built SaaS connectors.	Hybrid Deterministic Flows (state machines) + Generative Playbooks + Live-Agent Transfer.	Bridges into ADK when you need `before_tool_callback`, `ParallelAgent` / `LoopAgent`, or Git CI/CD gates!
How Data Stores & Tools Carry Over	Uses Agent Search data stores and Integration Connectors directly.	Uses Data Stores, OpenAPI tools, and CES fulfillment webhooks.	100% Data Store Reuse: Code-first ADK connects to the exact same Agent Search data stores or invokes the Studio agent via A2A!
Primary Architectural Trigger	“Business analysts,” “internal employee assistant,” “SharePoint/Jira connectors in days.”	“Inbound voice calls,” “IVR replacement,” “barge-in,” “live contact-center warm transfer.”	“Prototype validated in Agent Designer now requires custom Python callbacks / parallel workflows / Git CI/CD.”

4. Real-World Deep Dive: Data Store Reuse & Hybrid A2A Coexistence

Architect Question	How It Works Under the Hood in Production
“When we transition a prototype from low-code Agent Designer to code-first ADK, do we have to re-ingest and re-chunk our 50,000 SharePoint and Drive documents?”	Zero Re-Indexing Required! The Agent Search (`VertexAiSearchTool`) data store lives independently in the platform layer. Your code-first Python ADK agent simply references the existing `data_store_id`, preserving all document chunks, embeddings, and identity-aware ACL mappings!
“What if non-technical HR analysts want to keep managing the front-door employee assistant in Agent Designer, while software engineers need to add a complex multi-step payroll tax calculator built in ADK?”	The Hybrid A2A Pattern: Deploy the software engineering team’s ADK agent as an A2A Server (`to_a2a()`) and register its Agent Card in Agent Registry. The low-code Agent Designer assistant can then discover and delegate payroll tax calculations to the remote ADK agent over A2A while HR analysts continue owning the front-door experience!
“How does an Agent Designer assistant prevent a junior employee from reading restricted executive M&A folders stored in the same Google Drive data store?”	Query-Time ACL Enforcement: Agent Search ingests the source document permissions alongside the text. When the employee queries the assistant, Agent Search filters the vector/keyword index against that employee’s corporate identity before passing any chunks to Gemini!