How 11x Built Alice’s Brain to Make AI Sales Reps Smarter and More Human

Alice’s Brain: How 11x’s AI is Changing Sales Outreach

The use of AI agents in sales is evolving quickly, and 11x is at the forefront with “Alice,” an AI Sales Development Representative (SDR) that learns and operates like a human. Sherwood Callaway, Tech Lead, and Satwik Singh, AI Lead at 11x, shared their experience developing Alice’s knowledge base at the AI Engineer World’s Fair in San Francisco. Their work is simplifying and scaling the traditionally time-consuming sales outreach process.

What Does an SDR Do?

Sales Development Representatives focus on sourcing leads, reaching out, and booking meetings. Success is measured by the number of positive responses and scheduled appointments. Alice takes these tasks to a new level by sending around 50,000 emails daily across 300 organizations. That’s a huge leap compared to the 20-50 emails a human SDR typically sends. To manage this, Alice needs a deep understanding of both the seller’s products and the prospects’ pain points.

From Manual Input to AI-Driven Knowledge Base

Before, 11x sellers had to manually enter detailed information about their offerings, pain points, and value propositions into a dashboard to train Alice. This was slow and created friction for new customers. As a result, the emails Alice generated were either too generic or overloaded with irrelevant details, reducing the quality of outreach.

The breakthrough came with the Knowledge Base system that mimics how a human SDR gets onboarded. Instead of pushing info manually, Alice now pulls all relevant business data proactively. Satwik Singh explained that this process is similar to training a new human SDR by giving them documents to study over weeks or months. The Knowledge Base acts as a central repository where sellers upload all source materials in various formats—PDFs, images, websites, audio, and video.

How the Data Gets Processed

To make this diverse data usable, 11x’s system first converts non-text content into text that Large Language Models (LLMs) can understand. Rather than building this from scratch, 11x uses specialized third-party tools:

• LlamaParse for documents and images
• Firecrawl for websites
• Cloudglue for audio and video, extracting info beyond just transcripts

After parsing, the text is split into logical chunks and stored in Pinecone, a cloud-hosted vector database known for its efficiency and built-in embedding models.

Smarter Retrieval for Smarter Outreach

When Alice needs to answer questions about leads or craft personalized emails, it uses a “Deep Research” agent built with Letta. This agent takes lead info, formulates questions, searches the Pinecone database, and generates concise, context-aware answers. This multi-step retrieval process helps Alice respond with more relevant and nuanced messages.

Building Trust with Visual Transparency

To help sellers feel confident in Alice’s understanding, 11x created an interactive 3D knowledge graph visualization. Users can explore Alice’s “brain” by clicking on nodes representing data points and seeing the underlying information. This transparency reassures customers that Alice isn’t just sending generic emails but truly “knows” their business.

Lessons from Developing the Knowledge Base

• Retrieval-Augmented Generation (RAG) is complex and requires careful handling.
• Deploying in production early can provide valuable real-world insights, even before extensive benchmarking.
• Relying on specialized vendors for parsing diverse data types saves time and improves quality.

Thanks to these insights, Alice has evolved from a simple email-sending tool into an intelligent digital worker capable of human-like sales engagement.

For sales professionals looking to boost their skills with AI tools similar to Alice, exploring AI training courses can be a great next step. Check out Complete AI Training’s latest AI courses to learn how to integrate AI into your sales process effectively.