Top 5 Predictions for Cognitive Search in 2017

Following the shake-up in the industry caused by the announcement of Google winding down sales of the Google Search Appliance, 2017 sets up to be the year that enterprise search embracing the power of the cloud to finally deliver cognitive search.  After years of hype, the industry is poised to begin delivering the post keyword based search solutions.  Additionally, search will be the technology for processing large data sets for which the Big Data trend has emerged from within.

Everyone fundamentally understands the technology and the high level business case for it. The problem is that, few organization are agile enough to embrace it while  still fewer consultancies and agencies are capable of actually delivering a solution based on the technology.  As technology changes quickly, it is better to embrace the change before it is force upon you.  We plan on following up on the  trends with case studies and examples throughout the year.  It is going to be an interesting ride.

Given that, here are our top 5 predictions for Cognitive Search in 2017:

1. Google releases Springboard to open beta

Announced at Google Atmosphere Google Springboard, sets the stage for the GSA sunset as Google’s cloud based replacement for the Google Search Appliance.  Very little has been publically disclosed other than the mention of a Box.com integration at Box Works.  Springboard is Google’s first step in transforming from appliance-based solutions to “Search and Assist.”

While it’s not expected to be impressive in its initial release, neither was Google Apps.  Look for additional capabilities after its release on the “assist” side of “search and assist”  It will be interesting to see if they can achieve the promise of something like Google Now for the enterprise.

2. Cloud First - Major Vendors adopt SaaS-first model for new initiatives

Customers may prefer their data to stay under their control, but the cloud is the clear winner to provide the computing workloads required to perform the data processing to provide the natural language, machine learning and real time analytics that are becoming the mainstay of enterprise search.  Additionally, the cloud delivery model leads the way toward  easier deployment, faster release cycles, and simpler revenue models.

Look for more companies to simply offer indexing and analysis services rather than software to install and deploy on premise.  Open source will fill the void on-prem.

3. Machine learning transforms enterprise search to cognitive search

The benefits of understanding and extracting more signals from content will change  search from, “This document has the keyword in the title so it must be more relevant” to understanding user intent, observing behaviors and detection of other patterns to correctly assert the most relevant documents.

The added benefit of this is the reducing of the need to train the system through creating of boosting rules, adding of synonyms and tuning scoring.  This is an ongoing effort that most systems lack.

4. Analytics allows search programs become proactive and predictive

Systems like Coveo Reveal and Lucidworks Signals are examples of analytics determining relevancy.  Learning user’ behaviors with your content will allow  systems to identify the most relevant articles for that user at that specific time.  It is relevancy on a per user basis.

5. Microservices and Bots arrive in the enterprise

In order to help information workers and customers, assist agents will become more prevalent and integrated into users workflows.  This differs from the traditional monolithic search applications.  At MC+A we have evangelized search in the following way, “Don’t make search an opt-in experience.  Search should be where the users are; in their existing workflows.” This means doing away with the monolithic search deployment and focusing in on what users need and where they need it.

Migrating search into smaller more discrete areas cause cause high utilization of the service and technology and greater improvement in your workflows.