In today’s evolving digital landscape, businesses are inundated with high volumes of data.
This proliferation of information, while valuable, brings with it the challenge of effectively managing and retrieving relevant information at speed.
Integrating enterprise search capabilities that use artificial intelligence represents an incremental improvement as well as a game-changer, transforming how companies can access, analyze and activate their accumulated data repositories.
“The ability of generative AI models to be coherent and fluent when synthesizing material is a step function improvement in their output,” Eddie Zhou, head of AI at Glean, told PYMNTS during a conversation for the “AI Effect” series.
“Using these models as an interface to help businesses decide how to execute a set of tasks — that is a capability that was not present in AI models in the past,” he added.
Still, when it comes to the readiness of companies to use AI, Zhou underscored the challenges posed by the marketplace-wide silos of fragmented data landscapes, emphasizing the importance of clarifying use cases and understanding the value proposition of investing in enterprise search before deployment.
This is where AI steps in, armed with capabilities that range from natural language processing to machine learning, to not just search and sort data but truly understand it.
For instance, imagine an enterprise where employees, regardless of their department or role, can query a system in natural language, just as they would ask a colleague. Whether it’s a salesperson looking for the latest product specifications or an engineer seeking a specific piece of code, AI-based search tools deliver precise, relevant results — often anticipating needs before they’re explicitly expressed.
But capturing these benefits is not as easy, or as simple, as just flipping a switch.
“The data layer that drives a lot of knowledge work is messy, and many people within companies are underestimating the difficulty of the entire process” Zhou explained.
“A lack of clarity around what firms want to solve for can slow down integration,” he said. “Yes, AI can help, but what do you want it to do? It’s a moving target … we’re just starting to uncover where real value is added.”
At a high level, enterprise AI enhances internal search functionality by analyzing patterns and trends in data. The systems can offer recommendations, potentially uncovering valuable insights that might have remained buried in unstructured data. This not only enables operational efficiency but also allows for greater strategic planning and decision-making.
“It’s something that every knowledge worker, everyone who works on a computer, really, can get some value out of,” said Zhou.
“What AI search entails is figuring out where your company’s knowledge resides today, which is usually scattered,” he explained. “Different vertical sectors will have different specific places where their data lives, including documents they work in, communications they use, … and you need to connect it into one unified store, a search index, a knowledge graph. …You need this completeness in order to break down the silos, reach this promised land of knowing about and understanding what’s going on in other parts of the company that you would have never known otherwise.”
He stressed this needs to be done in a way that is permission-safe.
“One of the biggest differences between the consumer world of [large language model] applications and the enterprise world is that different people are allowed different access to different information,” Zhou said. “Permissions are a first-class thing you have to think about with enterprise AI.”
By indexing disparate databases and repositories, AI-powered tools create a unified search gateway. This accessibility can improve cross-departmental collaboration and knowledge sharing, fostering a more interconnected and informed workplace.
“Building these products in a transparent way that allows the user to understand what is happening is also key to building trust in the system,” he said. “…That’s why search and generative AI come together so neatly because search is traceable; it has provenance. And generative AI is a force multiplier.”
Looking to the future, Zhou envisioned AI systems that not only read and understand but also execute actions, freeing up cognitive resources for employees to focus on less mundane endeavors.
“I’m most excited about shifting the mindset … to how do I start to use an AI system that can actually do things beyond just answer a question,” he said.
“Search is just a sliver of every enterprise workflow, and you can see [future iterations of AI systems] expanding that sliver on both sides,” Zhou added. “I think a move toward these more agent-based flows that can both reason, read and do things is where I see a lot of the future headed in the next few — I was going to say years — but, really, months, when we look at the future for this space.”
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
