I’m a big fan of digital assistants and have an Amazon Echo device in almost every room. They turn on faucets, turn off lights, answer questions (though often not accurately), and I can even have conversations with them. But when it comes to applying them to actual work efforts, they fall woefully short.
This week, I came across a company — iGenius — that might be able to change that. It created Crystal to marry your vast data repositories with artificial intelligence (AI) to give birth to your very own enterprise-class digital assistant.
While iGenius is far from the only company working on this, Crystal is the best such tool I’ve seen so far.
Let me explain why.
The experience
Let’s start with the experience and the companies using this tool today, then explore how it works. You don’t get anything like an Amazon Echo; you interface with the tool using your phone or PC to enter a query using Crystal’s natural language interface in voice or text. It translates your question into an adequately formatted query against the connected data repositories (without copying the data) and, in moments, spits out a response. (That answer, depending on the question, may be just text or text and a graph.)
It will also automatically generate possible follow-up questions based on what you or others have asked before, so you can drill down to specifics you might need. It also retains context for further queries.
For instance, you can ask about sales by product category, and Crystal will deliver text and a graph that showcases sales performance for the current period. You can then specify, “No, I want year-to-date.” It should recognize you still want product category sales and provide an answer, while also suggesting you ask about underperforming products or other data cuts based on prior multi-dimensional KPIs (Key Performance Measurements).
This explanation is easier to understand with a demonstration showcasing a user interaction with Crystal and the results. This also showcases that Crystal can handle multiple languages and automatic drill-downs, which could be invaluable to an executive needing to answer questions from a customer or superior. It also can send an alert on pre-selected conditions such as costs or revenues that fall outside specific preset parameters. A sales rep on the way to a meeting can ask about, and get, critical background on a client before the meeting gets under way.
This case study shows how this tool was implemented to help Insurance agents working at Allianz. It functions in some ways like a sales assistant or secretary (when those positions were available to support agents).
That’s the experience; what’s happening on the back-end is equally fascinating.
Implementation
Typically, to create a tool like this, you’d need massive amounts of training. The first version of Crystal needed up to six months to integrate data, do the training, and have something that worked as expected. According to IGenius, with version 2, most of that work is automated — and the setup and training time can be measured in minutes.
This massive reduction in deployment time is the big differentiator with Crystal. Other enterprise-level digital assistant efforts have long set-up times, similar to what IGenius had initially. Having a tool like this, which takes only slightly longer than setting up an Amazon Echo, makes it a potential game-changer.
Crystal sits on top of your data infrastructure without altering it, so you don’t have to worry about breaking dependencies or most of the unintended consequences that can plague projects like this. Crystal only touches the data it needs for the query, limiting overhead and assuring minimal impact on operations.
However, getting to this point did require focus at iGenius; currently, the tool is only available for utility companies, energy firms, financial institutions, and the pharmaceutical industry. While administrator training on the first version of Crystal took months, that training time has shrunk to about a week, depending on the admin’s skill set. The implementation is generally dedicated, and like most new tools, is sold as a service with no capital expense. (iGenius generally works with a customer’s preferred consultant for rollout.)
IGenius has been around for about four years and the founder, Uljan Sharka, came from Apple, which helps explain a heavy focus on ease of use.
Wrapping up
AI’s advances into the enterprise have been fascinating — and generally disappointing because our expectations have been so incredibly high. iGenius is one of the few tools that might actually meet over-the-top expectations. It provides a natural-language interface to company data and uses its enhanced AI to not only answer the question you ask but the one you want answered. This answer discrepancy is often one of the biggest and most frustrating problems with query tools. To get what you want, you have to use the specific word order the tool recognizes.
That is the real promise of AI; it learns how to work with you, not the other way around.
