How Generative AI is Transforming Search
Generative AI is revolutionizing the search landscape by providing answers to virtually any question. As AI developers refine this technology, a new synergy between search and generative AI emerges.
Generative AI is reshaping three key aspects of search: how users ask questions, how data is sourced for answers, and how businesses deliver information to their customers.
AI Agents Are Here
For years, Google dominated the search landscape, capturing nearly 82% of search traffic. It dictated how users sought information and how brands appeared in results. Companies relied on search engine optimization (SEO) strategies, often crafting convoluted keyword queries. While this method produced passable results, users struggled to efficiently sift through countless website listings.
The advent of large language models (LLMs), especially as seen in chatbots like OpenAI’s ChatGPT, has changed the game. Users can now pose questions in natural language and receive immediate responses, eliminating the need to navigate multiple websites.
Christian Ward, chief data officer at Yext, succinctly encapsulated this shift: “An AI that understands search results and explains them to you is a monumental improvement over sifting through millions of pages.”
Natural Language Over Keywords
Generative AI encourages users to ask questions in everyday language instead of formatting them as keywords. This shift allows users to access the information they need without the hassle of searching through numerous websites.
Perplexity has embraced this evolution in search practices by positioning itself as a comprehensive search engine rather than just a chatbot for generating content. By partnering with data providers like Yelp and Wolfram Alpha, Perplexity has successfully enhanced its data-gathering capabilities, leading to increased traffic.
Recognizing this trend, Google has also adapted. During its Google I/O event, the company announced innovations that allow it to "do the Googling for you," including an AI Overview feature that summarizes query results in its Gemini chatbot.
For businesses, it’s not just about adjusting how they ask questions; they also need to ensure they can provide answers based on their specific documents.
RAG: A Game-Changer for Enterprise Search
Retrieval Augmented Generation (RAG) is gaining traction as a major trend in generative AI, offering model providers new ways to serve enterprises. RAG enables companies to ground AI models in their internal data, ensuring that results derive from their own documents.
Ben Flast, director of product management at MongoDB, shared insights on the growing interest in RAG: “While LLMs are advanced, we've seen a significant focus on customer support and internal use cases where companies are keen to manage risks effectively.”
RAG's strength lies in its ability to refer directly to actual documents, helping users find accurate answers.
Major players like Amazon Web Services (AWS) and Microsoft are already providing RAG-specific services, while companies such as Elastic, Pinecone, and Qdrant offer vector databases to support RAG frameworks. However, tools for monitoring RAG systems are still in early development.
As more enterprises adopt RAG, many of its applications remain internal due to concerns over AI hallucinations. Providers urge companies to thoroughly evaluate RAG models prior to implementation. AWS, for instance, has integrated RAG into its generative AI strategy through the Amazon Q product and introduced methods to test RAG result accuracy.
The Future of Company-Specific Search Platforms
As RAG technology matures, businesses may face a pivotal change in how they conduct searches. With increasing avenues for submitting queries, companies must decide whether to share their data directly or rely on information aggregators like Google. Taking control of their data presentation could lead to more tailored customer interactions.
Ward proposed the possibility of companies creating their own search platforms empowered by RAG and generative AI. This approach would enable customers to find information rooted in specific brand data. For instance, a consumer wanting to know the available colors of Everlane pants could ask the Everlane website directly instead of using a general search engine.
“It’s not the end of search, but rather a decentralization for specific queries. For general inquiries like locating the nearest pizza shop, Google excels, but for specific concerns, like allergen information, it’s best to consult the shop directly.”
Explore these themes further at the upcoming VB Transform 2024 conference, which will host expert panels on the future of AI. We hope to see you there!
Ranking
