Project · In use
Agentic icon-generation service
An anonymized enterprise service using Google ADK and Google AI capabilities to search, generate, and process icons against internal design guidance.
- Role
- Agent service design and implementation
- Context
- Enterprise delivery · Current role
- Status
- In use
Problem
Icon creation involves more than generating a plausible image. A useful enterprise service must search existing assets, decide when generation is appropriate, process the result, and keep outputs aligned with a design system.
Insight
The agent’s value comes from orchestration. Search, generation, processing, and guideline checks are distinct steps with different failure modes; combining them behind one prompt without structure would make the service harder to trust.
Decision
Build a tool-using agent service with Google Agent SDK and Google AI services. Give the workflow explicit search, generation, and processing responsibilities while keeping internal design guidance as a governing constraint.
Implementation
Austin built the service and connected the workflow stages. This anonymized case study does not disclose employer-specific prompts, assets, architecture, or design rules.
Evidence
- The service combined icon search, generation, and processing.
- It used Google Agent SDK and Google AI services.
- Outputs were aligned with internal design guidelines.
Reflection
Agentic design was useful here because the task naturally decomposed into tool calls and checks. The goal was not to make the workflow look autonomous; it was to make each decision legible and useful.
Evidence on record
What can be checked now.
- Combined icon search, generation, and processing in one agent service
- Used Google Agent SDK and Google AI services
- Aligned outputs with internal design guidelines