From Exploration to Building
1. From Exploration to Building
Imagine your sales director walks up and asks, why are we losing enterprise deals? Right now, someone has to dig through the CRM, read dozens of call transcripts, and piece together an answer. It takes hours, maybe days. What if an agent could do all that in seconds, search every conversation, query every metric, and give you the answer with actual evidence from customer calls? That's what you're building in this module. A sales intelligence agent that combines structured data and unstructured conversations to answer real business questions. In the previous module, you learned how agents are different from assistants. Agents plan, use tools, and work autonomously. That was the foundation. Let's build it. First, sales conversation transcripts. These are the calls your reps have with customers. Discovery calls, demos, technical reviews, negotiations, all unstructured text. Second, sales metrics, deal values, close dates, win rates, sales rep performance, all structured data in tables. The power comes from combining these data types. Remember your sales director's question, why are we losing these enterprise deals? Your agent searches through conversation transcripts to find objection patterns. It queries the sales metrics to identify which deals were lost, then synthesizes both into a clear answer with specific evidence from actual customer calls. That's what we're building. We're building this agent in Snowflake. We have two ways to build this, one with the API and the other using the UI. We will be using the UI, no heavy coding required. We'll point and click our way through the entire process. This means agent development is accessible while still giving you the production-ready capabilities. You can use the API if you need a programmatic way to create agents. Agents can use any number of tools to accomplish their tasks. They might query databases, search documents, call APIs, or execute custom functions. For this example, our agent will use two building blocks. First, we will build a semantic view for Cortex Analyst. This teaches your agent how to query structured data. It defines the business terms like win rate and maps them to actual table columns. Second, the search service. This lets your agent find information in unstructured text, sales conversations, customer emails, meeting notes. For the building blocks, we have our agent itself. This is our orchestration layer that decides which tools to use and in what order. We're going to build these in order, semantic view first, then search service, and the agent that ties them together. Each piece builds on the previous one. Within a couple of videos, you'll have a working agent answering sales questions.2. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.