Hopefully, a catchy title, but like it or not, Enterprise Search, a technology sector that has been around for decades, is rapidly and seemingly aggressively starting to leverage AI. On the surface, that should come as no surprise, as ultimately, making sense of and spotting patterns in vast amounts of random data is what AI does best. And at some level, Machine Learning and NLP (Natural Language Processing) have been used by Enterprise Search vendors for a long time. But things are nonetheless moving forward on the AI front at a clip.
Notice I use the phrase ‘Enterprise’ Search and not just Search or Search Engines? That’s because there is a world of difference between the search tools you use to surf the web and those you use at work. Web search technologies like Bing and Google are crawling across mountains of pages that have been well labeled and tagged. For example, the content on Nike.Com is figuratively screaming out, ‘pick me, pick me!’. Enterprise search engines, in contrast, are crawling across mountains of poorly, if at all, managed, labeled, or tagged content. It’s trying to find heavily disguised needles in giant haystacks.
The core Enterprise Search capabilities, be they from Seachblox, Sinequa, Coveo, etc., do as good a job as they possibly can in such difficult circumstances. The problem with Enterprise Search is seldom with the vendor; instead, it is with the buyer who expects miracles and cannot be bothered to invest the time and effort to tune and maintain good search results. AI can’t fix that, but it can help in refining search results and relevancy. AI can at least help to bring some order to the chaos, and it can help refine search results to the genuinely relevant ones.
If you want evidence of where things are going, French search giant Sinequa this week announced the launch of not one but four pre-trained deep learning language models for its platform. Four, not one – that is a vast amount of AI processing power. This means that the traditional statistical search, enhanced by NLP that Sinequa has long provided, can now be combined with ‘Neural Search’ capabilities. The goal is to improve the relevancy of any answers to natural language questions, with the deep learning modules making sense of context and overall meaning. In reality, today, overkill for many use cases, but it’s a big step.
The big lesson you learn as an industry analyst is that tech advances in enterprise software take a long time to gain traction. Often it takes five to ten years for widespread adoption to occur. At Deep Analysis, we focus on innovation and disruption in that new approaches coming from startups and established vendors alike in 2022 will likely take years before they truly take hold in the real world. Therefore, no matter how well Sinequa’s deep learning modules work, we don’t expect anyone to suddenly start ripping and replacing their existing Enterprise Search engines anytime soon. We hope to see, though, and will watch with interest over the coming year, some early adopters who, through trial and error, will find the right niches to exploit this new technology. Similarly, we expect to see the use of more advanced AI (think neural networks and deep learning) be explored more widely for Enterprise Search moving forward.
Nobody likes Enterprise Search; everybody wants it to work like Google. At the same time, nobody wants to do the work needed to make that a reality. So expect to see more vendors like Sinequa and Pureinsights add advanced AI capabilities to deal with the status quo and try to make Enterprise Search more loved, relevant, and accurate over the coming years.
If you are looking for new Enterprise Search technology for your organization, or you are a technology vendor that thinks it’s doing something innovative or disruptive let us know – our door is always open!
