Qlik Insight Bot - an AI-powered bot for conversational analytics built on Qlik’s Associative Engine

[QlikSenseUpdates]

"If something is made so simple that fools can use it, only fools will use it."

[QlikSenseUpdates]

Yet, in the field of business intelligence, making information more usable--more available, understandable, and digestible--is often an exercise in simplifying the experience. We summarize information because most raw data is too detailed for us to process. We distill huge amounts of complex data into simply displayed KPIs so that humans don’t have sift through rows and columns of data for insight. And we process data through machine learning algorithms to discover patterns that would take humans years to discover. All of these efforts are intended to make the process of generating actionable insights from data simple.

Human beings prefer what is simple.

Even business users with access to nicely designed but complex dashboards often prefer the simplicity of interfacing with a knowledgeable and delightful data analyst. It's simpler to type questions into a chat window than it is to navigate content from a BI portal, where oceans of data can feel like mountains of sophisticated analytic content. For everyday business users, it's easier to ask content creators for simple answers than to search for an answer like the proverbial needle in a haystack.

This is why at Crunch Data (acquired by Qlik earlier this year), we created we created CrunchBot.ai—an AI-powered bot for conversational analytics built on Qlik’s Associative Engine.

We built what is now called Qlik Insight Bot with the idea of providing conversational access to data insights as easily as conjuring trivia and music playlists from Alexa. We believe conversational analytics will be the key to delivering value to users who struggle to adopt analytics in general, users who prefer to work with data analysts as intermediaries, and even users who only sometimes struggle to grasp new analytic content.

I have spent years working with customers to drive value from data, first as a Qlik employee, and then as a co-founder of a small consultancy specializing in Qlik-related data engineering projects. It was clear to me that no matter how well an application suited a particular set of users, there would be another set of users for whom our solutions lacked the accessibility to deliver maximum value.

At one of my favorite companies, a giant multinational corporation known worldwide for its excellent software for small business and personal tax software, we built a set of Qlik apps on data from Hyperion and other sources, and we were broadly successful rolling out this sophisticated capability and content to the finance and HR groups. Both application sets worked beautifully for their respective audiences, with a relatively high degree of data literacy and domain expertise. But when providing access to a wider contingency, specifically to budget owners and people managers, we (predictably) found that broad simplifications were required to make the applications usable.

Budget owners just wanted answers generated in a simple manner:

• What is my budget?
• How much budget is remaining for this quarter? How about this year?
• Which employees are spending the most on travel?
• Show me spend for each sales rep.

People managers wanted to ask simple questions such as:

• What is my net promoter score?
• How does my net promoter score compare to the entire company?
• Which employees with a high retention rating are at risk for attrition?
• Predict attrition be over the next 6 months?

Predictably, it was easier for new users to get answers from their finance and HR business partners than it was to find them in an application built for Finance and HR.

Instead of asking developers to create custom views tailored to the specific needs and skill sets of different user groups, the advent of AI now allows organizations to offer a conversational analytics experience to complement visual dashboards, creating a new paradigm in which users can ask any questions about their data using natural language. We designed the Qlik Insight Bot to fulfill this exact purpose--to act as an analytic intermediary, in the same way that an analytics expert might intermediate on behalf of a casual business user.

The Qlik Insight Bot is now available, allowing users to type in natural language (NLP) questions consisting of every possible combination of dimensions, measures, and visualizations. It applies machine learning algorithms to find patterns, root causes, and anomalies, then responds in natural language (NLG). It frees the data analyst from constantly serving their less analytically inclined users, it and empowers the casual business user with an easy ramp into data literacy. Users can ask precise questions directly to the Insight Bot, and the Insight Bot responds in sub-seconds with answers and analysis that might take a human analyst several hours to create. Users have access to any answered available from Qlik Sense application, and they access it in the most natural, simple way possible.

This is not to say that simplicity is enough or desirable. Users rarely lack follow up questions, and a simple question and answer session with most any other "chat bot" available is a pale, half comparison to the conversational analytics enabled by the Qlik Insight Bot.

The difference is that other chat bots are limited to superficially answer questions. The Qlik Insight Bot engages users in meaningful dialog, inviting casual users not just to ask questions, but to discover and explore a depth of insight behind answers. When a user asks a question of the Qlik Insight Bot, getting a direct answer is only the beginning of the conversation. After delivering a concise answer, Qlik Insight Bot uses the Qlik Associative Engine and AI (Augmented Intelligence) to deliver a visual depiction of the answer based on what it has learned from users asking similar questions. It then provides a natural language (NLG) interpretation from machine learning (ML) algorithms. The pattern detection and causality inferences are embedded into the response, suggesting what (i.e. what products, employees, or external factors) could be causing changes in the subject performance attributes. Notably, we are also able to infer cause from what is conspicuously absent from a trend, this capability being uniquely possible with Qlik's associative engine, a with no analogous feature in competing products

But I argue, that even this isn't enough. The unknown pundit who pontificated that "if something is made so simple that fools can use it, only fools will use it" remains unknown for good reason, but he or she did have a point.

Conversational analytics as delivered by the Qlik Insight Bot are possibly the best way, and certainly the simplest way, of engaging users in meaningful data analysis; but ultimately, depth in understanding comes from personally exploring data visually and performing one's own analysis. Expediency and data literacy may prevent some users from immediately taking the dive into visual exploration, but it's important for them to have the option. Because the Qlik Insight Bot runs on the Qlik Associative Engine, which understands and retains context, the same engine drives both visual and conversational analytics and creates a seamless experience across them.  Being able to “bridge the gap”, users can enter conversationally, and the seamlessly dive into visual dashboards – in context – to explore further.  The Qlik Insight Bot to persistently augments the conversation with links to the underlying Qlik apps, gently reminding its users that further insight is just a click away.

Charney Hoffmann
Director, Consulting Offerings & Enablement