The way of discovering new information through traditional search engines using “search keywords” is fading away. Customer behaviour is shifting towards finding accurate information using a question whereby the information is a lot more tailored based on the customer’s needs. After the introduction of ChatGPT, customers are adopting this new wave of technology to its full potential. The Large Language Models (LLMs) that underpin ChatGPT have been evolving rapidly to cope with consumer trends.
Our product Document360 is a knowledge base platform catering to SaaS companies for building their self-service software documentation. We internally use our own product to build software documentation for Document360. Search is one of the critical features of a knowledge base platform as customers can easily find the right information instead of navigating to hierarchical structures of information. Given the latest shift in customer behaviour, customers are no longer interested in typing in search keywords to find answers! Customers now wish to enter a question via the “prompting” technique and seek accurate answers in the format they want! This benefits the customer in multiple ways, such as
- Getting accurate answers
- Getting a holistic response in a short span of time
- Understanding the contextual information
- Ability to fine-tune questions based on responses
To adapt to the emerging customer trends and take advantage of technological capabilities, today we are excited to announce the launch of our AI Assistant, Eddy. Eddy will provide answers to customers’ questions utilizing the LLM technology based on the knowledge base project it is pointed to. We will soon roll out our AI Assistant Eddy to our existing customers.
Our goal here is to create a conversational, human-centered search. We want to make it possible for users of our customers’ knowledge base to receive instant, trust-worthy, and accurate solutions to problems using conversational search powered by Generative AI capabilities – says Saravana Kumar, CEO, Document360
There is a lot of innovation that went into building this AI assistant.
Architecture
A reference architecture has been proposed which involves multiple components. This proposed tech stack for finding solution to our use case is complex and investment heavy. Moreover, having your knowledge base articles scattered across multiple platforms create “data-syncing” issues. To address this, we have utilized cutting edge capabilities from our existing vendor, MongoDB re-architected our solution as shown below. This architecture is robust and cost effective! We collaborated with MongoDB who offered technical assistance in optimizing our solution design and their vector search engine is blazingly fast!
Design
Instead of providing two search boxes for our customers based on traditional search and AI assistant, we combined both in a simple interface. This helps with the change management aspect of technology shift. This design aligns with our design philosophy of having human-centric design! This ethos is embedded in all our design practices across our product portfolio. Our customers can switch between different search engines depending on their need. Moreover, our customers will have an option to set any of our offered search engines as default for their users.
Experience
Our AI assistant is designed to cater to needs of our customers’ queries. Only answers from the knowledge base relevant to context of the question will be provided. The AI assistant is security hardened for prompt injection attacks and follow ethical guidelines in answering the questions. To ensure trust, all answers are cross-referenced with relevant articles and shown to the users. This enhances the user experience greatly!
Our journey forward
Building this AI assistant required an incredible effort from our team involving UX designers, product managers, developers, solutions architects, and other stakeholders. We have run back-to-back-to-back sprints, pushing ourselves to our limits. With the news of our roadmap now out in public, it’s time to begin the marathon of bringing these exciting new AI-powered tools to users and customers, listening to feedback, iterating, and improving. We’re excited to learn from our community and customers as we evolve our AI Assistant Eddy for the next era of technology.
