The discussion around generative AI has exploded in recent months. At Lucidworks, we are excited about the new opportunities presented by generative AI and large language models (LLMs), and we have invested and will continue to invest in experiments to identify where we can continue to deliver the most value.

Our goal has always been to enable amazing search experiences that empower both customers and employees. We’ve been a proponent, user, creator, and trainer of various language models for several years. As these technologies advance, we’ve applied them where they deliver value and show a new opportunity to improve the search experience. In addition, we understand the desire for the more conversational solutions that LLMs enable.

Ecommerce companies need AI-based solutions

Our own research at Lucidworks has made something very clear – AI, and especially generative AI enabled by LLMs – is absolutely a necessary element of the future of ecommerce. Additionally, search practitioners universally agree that AI can improve their ability to do their jobs and innovate. 100% of respondents to our search relevance survey ranked search relevance as highly important, with 96% saying it is difficult to deliver. A whopping 88% believe AI will be important in delivering relevant search in the future.

Given this, it’s no surprise that many want to jump in with both feet. That said, it’s important to ground this excitement in realism in order to best utilize its value. To cut through the noise, we put together a quick field guide outlining the need, potential, and cautions of LLMs.

A new field of potential

Many companies are moving fast to modernize their infrastructure in order to match the demands of an ever-changing ecosystem. Customers are opting to work with companies that have a digital strategy that meets the needs of the market.

This is especially true in the world of search relevance for ecommerce and knowledge management systems. Search practitioners expect generative models to transform the discovery experience in several key ways, including:

  • Conversational search and browsing that can respond to simple requests, such as “show me more in the color red,” with detailed results
  • Semantic search that becomes smarter by leveraging LLMs to produce stronger training data
  • Contextual document summaries that relate to the user’s story
  • Personalized product detail descriptions that highlight unique product attributes relevant to the users recent interaction

Opportunities to explore

So we know the need is there, and there is a ton of potential. As search experts, here’s what gets us most excited about the potential of LLMs:

  • Hybrid search and LLMs form a “virtuous cycle,” grounding responses in fact
  • LLMs improve semantic models
  • Retrieval augmented generation grounds LLM responses in fact
  • Conversational discovery
  • LLMs enable personalized, information dense responses
  • Document summaries
  • Explanations
  • Suggested filters
  • Revenue-driving recommendations and personalization

Hundreds of generative AI and LLMs are already crowding the space. The important factor to consider is that many of these applications are relatively new and their commercial application is both unknown and untested.

Search experience providers have an important role to play in exercising due diligence to ensure the cost-benefit and fit-for-purpose of models for specific tasks.

Where to be cautious

As exciting as this new field is, there is always room for grounding. Many companies are announcing integrations with generative AI applications (such as OpenAI’s ChatGPT and other LLMs) without explaining how the responses generated will be grounded in facts. There is also little discussion around how up-to-date information will be incorporated into responses.
In other words, the market is in hype phase, creating a lot of pressure to “announce something” relating LLMs to ecommerce and knowledge management platforms. There are numerous areas yet to be explored and tested when it comes to generative AI and search. They include:

  • Pricing
  • Result accuracy vs. hallucination
  • Data usage, security, and ownership
  • Rights
  • Content suppression or exclusion
  • Document-specific security and entitlements
  • Workflows
  • Processes for reviews and approvals

There is good reason to be excited about the potential for generative AI to transform the search experience, however it’s extremely important to see through the noise in order to avoid pitfalls when experimenting with and adopting new technologies.

Adopting guardrails for safety

For practical applications to work, AI often needs additional technologies to be attached to, or append to the workflow. Search experience platforms are an ideal environment to apply tech that grounds LLM responses in fact. Lucidworks’ Fusion platform is an example of an accelerator and protector technology that can help bring AI to search the right way. The platform provides the data enrichment, facet controls, security trimming, recommendations, and merchandising intelligence needed to deliver transformational AI capabilities that create meaningful experiences in a practical, trusted, and useful tool.

Interested in learning more about how Lucidworks Fusion can aid in navigating the budding LLM landscape? Get in touch with us today.

About Eric Redman

Read more from this author


Contact us today to learn how Lucidworks can help your team create powerful search and discovery applications for your customers and employees.