Inspiration
I have lost the count of times I have visited establishements like restaurants, salons, hardware shops etc just by chance and left mighty impressed. What surprised me most in these situations was the fact that there wasn't any mention of them when I conducted my diligent research asking for suggestions to three chatbots at a time! Local businesses often have deep community support that is visible in ratings, reviews and repeat custom, yet that reputation may not appear when customers ask an AI assistant for a recommendation. Agencies that specialize in SEO and website maintenance can help close that gap, although identifying the right businesses usually requires separate research across maps, websites, review platforms and AI answers. Goldmine brings those signals together in a focused prospecting workflow.
What it does
Goldmine searches a selected area (piloting with London) and business category, builds a Google Places candidate cohort and compares customer reputation with AI discoverability. The prospect detail page shows rating and review evidence, the ten named customer queries, Gemini responses, AI visibility, Gold score, website evidence and an outreach draft. The Gold score highlights businesses with credible local customer proof and a meaningful AI visibility gap. It is calculated from reputation and the measured gap in the completed query set. Businesses with positive evidence of multiple local locations are excluded from outreach.
How we built it
The product uses a TypeScript Vite frontend and an Express backend deployed to Google Cloud Run. Google Places API (New) supplies candidate and reputation data. Gemini on Vertex AI answers ten grounded customer-style queries for each measured prospect. Firestore stores runs and evidence. Cloud Build deploys GitHub main to Cloud Run. PageSpeed Insights provides selected-business technical evidence. The workflow keeps the initial scan fast by separating market discovery from selected-prospect enrichment, brand verification and outreach generation.
Google technology used
Goldmine began as a prototype in Google AI Studio and is deployed on Google Cloud. It uses the Google Gen AI SDK with Gemini 3.7 Flash through Vertex AI for grounded AI-visibility checks and evidence-based outreach drafts. Google Cloud Run hosts the application and Cloud Firestore persists scan runs and prospect evidence. Google Places API provides the local-business candidate cohort, while PageSpeed Insights API supplies selected-business website evidence. Cloud Build deploys the public GitHub main branch to Cloud Run.
Challenges we ran into
The product had to balance fast interaction with evidence quality. A Google Places search returns a candidate cohort, so the interface states that it is not a complete market census. AI visibility needs repeated, visible queries rather than a single opaque answer. We also had to move expensive per-business tasks out of the initial scan, while keeping the prospect detail useful and traceable.
Accomplishments that we're proud of
A live Google Cloud workflow that runs from local candidate discovery through to outreach. Ten named, inspectable customer queries for every complete AI visibility measurement. A transparent Gold score that connects customer reputation with an AI discoverability gap. A prospect view that turns research evidence into an actionable outreach draft. Clear handling of incomplete evidence, candidate-cohort limits and multi-location exclusions.
What we learned
An agentic prospecting tool earns trust when its claims are inspectable. Showing the actual queries, responses and calculation makes the result easier to assess and improves the resulting outreach. We also learned that latency is closely tied to product design, since a useful scan requires a deliberate boundary between immediate evidence and deeper selected-prospect work.
What's next for Goldmine
The next release will extend structured London-area coverage, add scheduled refreshes, strengthen contact and website enrichment for selected prospects, and introduce paid access for additional verified outreach drafts.
Built With
- cloud-firestone
- gemini
- github
- google-ai-studio
- google-cloud
- google-places
- googlecloudrun
- pagespeedinsights
- vertexai
Log in or sign up for Devpost to join the conversation.