Every app needs GenAI search - that’s why we’re giving it away for free

5 min read

Last edited:  

Every app needs GenAI search - that’s why we’re giving it away for free
Chris Naco
Chris NacoProduct Manager @ DevRev

Search is everywhere.

You can find an unassuming-but-powerful search bar in just about every product you use on a day to day basis. Search is a core workflow in use-cases ranging from e-commerce, music streaming, messaging, and work management. Search ranges from a convenient feature that saves you time, to the whole product (think Google).

Search is broken.

Traditional search - particularly in enterprise use-cases such as digging through work items or querying a help center - often fails to deliver a good user experience. The standard approach in these cases is to return a list of links that users must sift through to find what they are looking for. These items may be properly ordered (mostly) but are unlikely to be aggregated, summarized, contextualized, or personalized. Great search algorithms are difficult to get right. But every algorithm places the burden of finding relevant content back on the user.

This process leads to abandonment and frustration. The reliance on the even more outdated keyword matching can only worsen these issues, as it often fails to grasp the context or intent behind a query. The cost of abandonment is hidden but very real. Every user that fails to resolve their question, fails to increase their usage of the product, and ultimately the value they receive and reliance on your product.

Search is more important than ever.

Today, there is more and more data to sift through - either produced by people, machine logs, or LLMs. Google raised the bar on search - highly accurate and blazing fast results - while LLMs are changing the game entirely. With consumers demanding faster engagement and better experiences, you can’t afford to have a sub-par search lest you lose the attention of your user.

AI search is transforming how websites and apps interact with their users. Unlike traditional search methods that often return a list of possibly relevant results, AI search utilizes AI to understand and respond to user queries with precise and relevant information. This evolution in search technology is not just a trend; it's increasingly becoming table stakes for interacting with modern enterprise productivity tools.

Generative AI + observability = best in class search experiences

AI search dramatically enhances user experience by personalizing search results based on what the user is asking, what they have been doing and how they use the product. This is all made possible by combining Generative AI with observability. Combining these technologies allows AI search to deliver precise and contextually relevant answers. By directly providing the information users seek, AI search reduces the need for multiple search attempts, cuts down on user frustration, and speeds up adoption. This not only boosts user satisfaction but also increases the likelihood of deeper engagement and adoption of the platform. By delivering a more seamless experience, businesses are able to reduce costs (by lowering Support Tickets) and increase the lifetime value of users by improving user engagement .

How AI search works

AI Search runs on what is called a RAG pipeline. At a high level, a RAG pipeline is very simple. When put in practice, the details are what make it difficult to do well. Here is how one works:

  1. Take all of your data and break it into chunks.
  2. For each chunk, calculate the embeddings on it - imagine you could represent the chunk of text in thousands of numeric variables.
  3. Now take your Search Query and do the same. Calculate the embeddings.
  4. Of all of your chunks, find the most similar (imagine you’re back in high school geometry using the law of cosines). This retrieves the chunks or articles you would normally want to return in a search.
  5. Construct a prompt based on the nearest chunks to feed your LLM.

You can see where it quickly becomes complex. Where and how much do you chunk your data? Do you build your own embeddings model or use an off the shelf model? How do you build your prompt chain? For a deeper dive on how our search and embeddings model works, you’ll find it here.

Implementing AI search with ease

Integrating AI search into your website or app is straightforward with platforms like DevRev. DevRev Search automates the complex tasks of scraping help center data and performing prompt engineering. Businesses can easily add AI search capabilities with just a few lines of code, significantly reducing the barrier to entry for implementing advanced search technologies. The profile of companies that we think will find the most value of what we have built would be

  • A mid sized software business looking to quickly launch a GenAI feature and begin experimenting with how their users engage with it.
  • A small startup that wants to help their users to get the most out of their products, but are not yet ready to scale a support team.
  • A small ecommerce business that wants to be able to answer simple questions for their customers to never miss a sale, without having to spend all day answering those questions.

Implementing AI search is easier than many might expect. Companies like DevRev offer solutions that minimize the technical challenges, making it accessible even for those with limited coding expertise. Starting with AI search involves understanding your needs, choosing the right platform, and implementing it to begin reaping the benefits of enhanced user interactions.

This is part 3 in a series of blogs, case- study, white-paper and podcasts on “AgentOS & its capabilities". Stay tuned…

Chris Naco
Chris NacoProduct Manager @ DevRev

Builder passionate about all things AI and product