God made man,
Man made machine,
And machine made superman.
The past five decades have shown us how software can dramatically increase human productivity. In our consumer lives, we’ve seen search become the dominant metaphor for how we interact with information - whether it is searching for a Thanksgiving recipe on Google, for Cyber Monday deals on Amazon, or for the nearest Uber. I challenge you to find someone who hasn’t searched for anything today.
Yet ask the same question of your colleagues searching their enterprise data, and you’ll get a different response. Adoption of enterprise business intelligence solutions has lagged behind consumer software, mostly due to how difficult they are to use. For example, most business intelligence products require weeks of training for anyone to really learn how to get value from them. This creates job security for the few that get trained and guaranteed revenues for the software makers. What’s left unaddressed is the request for easy access to data from millions of knowledge workers.
Removing the middleman and letting people access their data directly requires two fundamental changes to status quo:
User interfaces need to become a lot simpler to push adoption from a few hundred thousand to tens of millions of users.
Man and machine need to work more in concert to deliver accurate answers, instantly.
There have been multiple attempts at the latter in the recent past, notably from natural language processing (NLP) solutions like IBM Watson. The shortcoming of those solutions comes down to the months required to train semantic search models for each vertical, and the probabilistic nature of results - even the best NLP solutions offer only about 80% accuracy.
What about the adoption piece of the puzzle? Search interfaces have solved the challenge of offering an easy way for man and machine to interact. Just as with Google typeahead suggestions, the machine can offer suggestions as someone types in a query, and this can be refined in real-time to arrive at the exact question the user has in mind. At the same time, the algorithms running those queries can benefit from iterative learning from human interactions. This will teach them what data is accessed more frequently, or what types of queries are run by different user personas. Combining this with the algorithm's ability to understand the properties of the data and how they relate to each other will help refine its suggestions back to the user.
A consumer-like search interface may not work perfectly for every enterprise scenario. While it might be ideal for me to get a page full of links back from Google when I ask for recipes, I want only one answer when I ask my business intelligence solution for “revenue last year.” This calls for a new breed of search engine - one for numbers - that can calculate accurate results and charts based on a query typed by the user.
This new breed of search engine for numbers has the potential to open up access to data across your entire company. Join ThoughtSpot and Palantir in an upcoming webinar and learn how to:
Move beyond the “create vs. consume” dichotomy by giving all users a familiar tool to interact with their data
Define useful filters and parameters for an internal search-driven analytics interface
Govern your data centrally but democratize access to all through search-driven analytics
Sign up here.