Sales reps dedicate just two hours each day to active selling, according to HubSpot research. At Snowflake, our sales team found they were wasting 10 to 15 minutes searching for the right content every time they needed to answer a single question, like “Can you explain how Snowflake handles data integration from various sources?” Valuable content was scattered across different platforms, forcing employees to hop between various tools to assemble the right information. The process was often cumbersome and inefficient.
To help them cut through the noise and have more time for selling and managing relationships, Snowflake developed a chatbot — the Knowledge Assistant — that can answer crucial questions with immediate answers it gleans from our sales enablement tool, our documentation site and more than 800 customer stories from marketing, value engineering and professional services. Now if a customer asks, “Do you have any successful gen AI customer stories?” sales reps can provide a real-time answer on the call rather than promising to look into it and get back to them later.
Built on our own Cortex AI technology, the Knowledge Assistant enables quick, conversational insights to expedite any research element of the selling process. It is a centralized place where salespeople can combine their sales acumen with all of Snowflake’s resources in seconds. Salespeople can ask a general knowledge question such as “What is the value of Apache Iceberg™ tables?,” request assistance in drafting initial client pitches, run an account analysis, help with meeting preparation or create tailored client pitches — and it’s all powered by generative AI.
The Knowledge Assistant uses Cortex AI for its search power, open source LLMs for parsing the information and Streamlit to distill and present the content in an approachable way. It maintains accuracy in pulling relevant information and provides detailed support for its answers with a direct link to the source. The bot is equipped with guardrails so that if the question can’t be answered, or if it doesn’t understand the question, the chatbot will explain that it can’t provide an answer, rather than hallucinate.
Key features of the AI-powered Knowledge Assistant
The Knowledge Assistant streamlines the sales process and empowers our teams with near-instant access to crucial information. By combining advanced AI technology with our extensive knowledge base, we’ve created a tool that’s both powerful and user-friendly — a new interface layer between humans and their data.
Here are the key features that make the Knowledge Assistant an indispensable asset for our sales organization and suggest what the future of sales holds:
-
Unified knowledge base: The assistant accesses materials from our sales enablement platform, customer stories, docs.snowflake.com and other sources, eliminating the need for users to know specific file locations.
-
Natural language processing: Users can simply ask questions, and the chatbot retrieves relevant deal-related information or facilitates deeper exploration through follow-up questions.
- Comprehensive support: The tool spans the entire customer lifecycle from prospecting to customer success, including conducting customer research, crafting relevant emails with context, assisting during negotiations, providing support for customer success and upselling, and answering technical questions.
- Human in the loop: While it’s AI powered, we never expect or intend to have the tool respond to customers automatically. There will always be a human vetting process, ensuring the accuracy and appropriateness of responses and helping ensure there’s always white-glove support for our customers.
Saving sales’ time one question at a time
The Knowledge Assistant is built on a combination of Snowflake’s proprietary technology and open source tools. While Snowflake is renowned for its structured data support, this chatbot is a perfect example of what Snowflake can do with unstructured data such as emails, PDFs or Google Slides, which don’t follow specific formats or logic. On the backend, Snowflake handles all the information indexing and retrieval, while open source tools handle the data parsing and chunking. Using Cortex Search and Cortex LLM functions, the Knowledge Assistant ensures near-immediate responses for the sales team presented on Streamlit.
The tool is already saving that 10 to 15 minutes per question by giving individual salespeople almost immediate answers. As the tool evolves into a collaborative platform for sales engineers, account executives and SDRs, we anticipate it could potentially save up to 40% of the sales team’s time.
Beyond sales, we expect to repurpose this chatbot for other business use cases, such as the marketing and legal departments, to provide similar value.
Using Snowflake to sell Snowflake
By building this chatbot internally using Snowflake’s technology and open source tools, we’re not only solving our own challenges but also demonstrating the platform’s power and flexibility. With Snowflake, companies can develop sophisticated AI-powered tools tailored to their specific needs.
As we move forward, we’ll continue to evolve this sales assistant, incorporating new features and capabilities to meet the ever-changing needs of our sales organization and, ultimately, to better serve our customers.
Ready to build your own AI-powered sales assistant? Visit our docs to learn more, including getting-started guides and limitations.
