SemiMassive • 2025
Razor Agent | Customisable Business Intelligence Assistant
A customisable Business Intelligence Assistant that turns your company data into an AI-readable format. Clients add their own internal data safely and use the agent in ways general AI tools can't, because it's built around your business, not public data.
Problem
Sales teams spend hours manually researching prospects, copying contact details from websites, and building lead lists by hand
Intervention
AI-powered web crawling with structured extraction, RAG-grounded chatbot, and unified Sales Assistant mode
Result
Automated prospect research with enriched Excel/CSV exports; AI assistant for outreach and competitive positioning
Human Impact
Zero headcount replaced; automated prospect research freed the sales team to focus entirely on relationship-building and closing.
Portfolio Venture 02
The Prospecting Bottleneck
Sales teams in industrial B2B spend hours every day on manual research. They visit a prospect's website, scroll through pages, copy-paste names into spreadsheets, hunt for project details, and try to piece together who does what and where. By the time they have a usable contact list, the opportunity window has already started closing.
The intelligence exists on public websites. The problem is that extracting it is slow, tedious, and inconsistent. Every rep does it differently, misses different details, and formats it their own way. The result: scattered data, missed connections, and wasted selling time.
From Manual Research to Automated Intelligence
Razor Agent is not a chatbot. It is an industrial intelligence platform.
Enter a target company's website, and Razor Agent deep-crawls it, extracting structured profiles of key people, active projects, office locations, services offered, and commodities handled. Every result is tagged by source reliability, deduplicated, and stored in a database ready for Excel/CSV export with enrichments like project type, firm role, commodity sector, and geography.
The Technical Stack
Led by Nick Strine, SemiMassive engineered Razor Agent as a full-stack intelligence platform, not a wrapper around an API.
- Frontend: React 18 + TypeScript + Vite with Shadcn/ui components
- Backend: Python FastAPI with Express.js proxy layer
- AI Layer: Google Gemini for generation and structured extraction, gemini-embedding-001 for semantic search
- Vector Store: Pinecone serverless for RAG retrieval
- Data Store: PostgreSQL for structured research data and conversation history
- Extraction: Trafilatura for web content, pdfplumber + Gemini Vision for documents
The architecture separates platform capabilities from client configuration, enabling rapid deployment of customised instances.
"Before a business can speak to AI, its data must first learn to read. The value of an agent is the proprietary data you give it."
— Nick Strine, Founder, SemiMassive
Who It's For
Razor Agent is built for B2B companies in mining, energy, infrastructure, and engineering services, anyone whose sales teams spend significant time manually researching prospects and building contact lists.
- EPC contractors mapping project opportunities and key contacts
- Equipment suppliers identifying decision-makers at target firms
- Consulting firms building competitive intelligence on specific sectors
- Trade associations mapping member ecosystems and market structure
The subscription-based model suits both mid-size firms scaling their prospecting and large enterprises looking to automate research across multiple teams.
Current Deployment
The platform is live with Beaver Process Equipment (Australia), who validated the research tool and are actively using it for ongoing, ad-hoc company research. The chatbot is configured with Beaver-specific personas, documentation, and competitive intelligence.
Next Phase
- Unified Sales Assistant connecting research data to the RAG chatbot
- Multi-tenant architecture for client isolation
- Usage metering and subscription billing
- Admin dashboard for self-service configuration