GTM tools are everywhere these days. Most follow the same pattern: pull signals, do enrichment, send outbound. We are big fans of Clay here at Pinecone because of its intuitive spreadsheet interface.Why Clay WorksExcel has always been intuitive for most people. You're essentially programming through rows and columns acting as control flows (e.g., while or for loops), cell references as variables, if formulas for conditionals, array or lookup functions as primitive data structures, other formulas as basic math libraries, tabs as subroutines, etc (fun fact: Excel is actually Turing complete). Clay takes the basic spreadsheet, pulls in a ton of data sources, adds in some AI functions, and automates GTM processes (e.g., sales outbound, lead scoring, CRM enrichment, account research, etc). Non-coders can build complex processes without technical barriers, otherwise known as GTM engineering.We use Clay extensively at Pinecone. The interface simplicity makes it easy to look up information, enrich data, and send messages through tools like Gong, as an example of a GTM outbound process.The Problem: Generic OutreachMost automated outbound emails were rather generic without much context to the company. It is time consuming and manual to research every company, explain how our products fit their specific needs, and include relevant customer references. We wanted to send more personalized emails that can give this relevance. Can we combine Pinecone and Clay into a workflow that automates this research process? Yes.The Solution: Semantic Search Powered OutreachHere is how we did it at a high level.The core part of this solution is a FastAPI service that orchestrates all the components and exposes an API interface for Clay to call. Inside the service, there are a few core components.A case study retriever that retrieves case studies from a Pinecone vector database. Pinecone indexes every one of our customer case studies. For example, for a financial services company, the system retrieves the Vanguard case study; for AI agent companies, it finds the CustomGPT or Delphi case study.The web crawler searches relevant company news related to AI initiatives.The email copy generator takes the case study and company news and passes everything to a large language model (LLM) to generate a personalized email.The service is deployed to Google Cloud Run, and Clay is wired to the service via the HTTP API action. Rows of company names trigger the service which generates the body of a personalized email. This in turn is passed to Gong or any other email system for output.💡 Note: the code shown is slightly shortened for the purposes of the blog post and may differ from the code repo.Below we will walk through some of the code. This system obviously required some coding, but it's easily reproducible in a low-code tool like n8n.Web Crawler Part 1: News Aggregation and Content ExtractionThe web crawler searches for company news using Google Custom Search API. A Google query is constructed based on key publications reporting on the company and the company’s own domain. This is fed to the Search API with only results from the last year returned.def get_company_news(company_name: str, category: str ="technology", max_results: int = 10):