Sourcy’s multi-agent platform helps turn product ideas into manufactured, quality-controlled products with Google Cloud.
Sourcy’s multi-agent platform helps turn product ideas into manufactured, quality-controlled products with Google Cloud.
Closing the idea-to-product speed gap
with multimodal AI
Consumer brands increasingly compete on launch velocity. Social and marketplace trends move in days, while traditional product development still takes months. The work begins with rough inputs from buyers and continues from early specifications through to delivery. Converting inputs into a manufacturable product requires manual expert coordination, making the process slow, inconsistent, and difficult to scale.
Sourcy is building a multi-agent, prompt-to-product platform to close that speed gap. Its agentic workflows convert customer intent into supplier-ready specifications. By discovering and engaging suppliers, Sourcy helps prepare comparable quotes to ensure buyers are getting the best possible deals. Familiar production workflows are AI-assisted but still managed by Sourcy’s experienced operations team. Google Cloud gave Sourcy the multimodal models and unified AI, compute, container, and database stack needed to process this workflow across languages quickly.
Gemini lets us turn a buyer's rough description into a factory-ready specification in seconds. We process around four billion tokens per month on Google Cloud, and the intelligence layer is still only one-fifth of our bill. That economics is what makes AI-native sourcing actually work.
Awsaf Alam
R&D Lead, Sourcy
Scaling prompt-to-product sourcing with agentic AI
Sourcy built a multi-agent sourcing system on Gemini Enterprise Agent Platform (formerly known as Vertex AI). Sourcy Brain orchestrates specialized agents for customer activation, specification building, supplier discovery and engagement, and quote preparation. The system uses five model families through the Gemini API, routing high-volume tasks to Gemini Flash Lite and complex reasoning to Gemini Pro. This model tiering has supported billions of processed tokens at a blended AI cost of roughly $1 per million tokens.
"Gemini lets us turn a buyer's rough description into a factory-ready specification in seconds," says Awsaf Alam, R&D Lead at Sourcy. "We process around four billion tokens per month on Google Cloud, and the intelligence layer is still only one-fifth of our bill. That economics is what makes AI-native sourcing actually work."
Deploying these services on Cloud Run also helps the team scale while continuing to manage a heavy transactional load. Sourcy was able to grow from two services to 23 with the company’s existing team size. Hosting these services in Google Cloud's Singapore region also keeps latency low for its Asia-based supplier workflows. Each workflow also captures current supplier and quotation data for future reuse, building a proprietary sourcing-intelligence layer.
Sourcy is a multi-agent prompt-to-product platform that helps consumer brands turn ideas into manufactured, quality-controlled, and delivered products.
Industries: Technology, Startup
Location: Singapore
Products: Gemini Enterprise Agent Platform, Gemini API, Compute Engine, Cloud Run, Cloud SQL, Google AI Studio, App Engine, Document AI, Speech-to-Text, Firestore, Pub/Sub, BigQuery
