There is no dearth of business “intelligence” tools. So why do up to 80% of corporate BI projects fail, according to Gartner research?
Lakes full of data, millions of dollars invested in dozens of tools... but yet not a drop of insight. Why? Do we as business professionals just like to guess our way through life? Does our company not have a data strategy? Do we have a process & culture issue?
The answer doesn’t just lie in poor processes or company culture — it’s in the tools themselves. Most classic analytics tools that produce visualizations were built for the simpler days: when Small Data was hip, it was acceptable to wait months for analysis, and confirmation bias was cool. But the world has moved on. A digital business of today needs to connect with its customers in a whole new way, faster than ever. And to do that, you need intelligence at every point of decision-making.
AI puts the “I” in BI
We are now in the Fourth Industrial Revolution, where everyone and everything is being redefined by intelligence. Take the analytics dashboard, for instance. For a generation, analytics dashboards were clumsy, monolithic and mostly told us what had already happened. Today, that’s no longer the case. With the help of AI, intelligence has revolutionized the humble dashboard, elevating it to a whole new level as a strategic tool. In this new, transformed intelligent experience:
- You are no longer searching for data; the insights find you.
- Insights are predictive, guiding you on what may happen and what action to take.
- Insights and recommendations are infused within your workflow and business context.
- Insights are delivered in a narrative format that provides greater context about a situation and possible results of your action.
AI works for you
AI (automated machine learning specifically) powers the intelligent experience. It’s about making things simpler and augmenting an employee’s skills and domain knowledge with precision guidance so they’re more productive and more successful in their day-to-day tasks.
So, how do does AI augment the analytics workflow?
- AI accelerates the path to discovery. The biggest friction in analytics is clean data. Why? Because data is stubborn. It doesn't like to move. It’s hard to keep clean. But with machine learning, data preparation algorithms can quickly run a profile of your data (millions & billions of rows) and present you with opportunities for clean-up (for example, a number in an alphabet column), data completeness (predicting missing values for certain rows), modeling (suggested joins of fields that seem similar), or enriching (understanding a related dataset). Think faster and more automated data preparation that gets you to better insights, faster.
- AI tells a story with your data. Humans are not capable of sifting through millions of rows of data to discover connections and patterns. AI, on the other hand, was literally built for that. It can sort through millions of rows almost instantly to unearth the most statistically significant patterns in your data. Furthermore, AI can provide an explanation of why a pattern exists, how different actions could impact that pattern, and what might happen in the future. You may see that sales of a certain product are decreasing, but AI-powered analytics can tell you the context around it, like the “why” and what other factors are affected by decreased sales. This is what the industry calls automated data discovery.
- AI delivers precise insights and recommendations to guide you. AI can provide insights that are not just reflective, but also prescriptive, so that you as a business user can make a decision that impacts an outcome. AI transforms analytics into a personal guide for your role or business that allows you to serve customers better. For example, an AI-powered analytics tool for a financial services company can analyze current customers’ saving habits and lifestyle data to better target potential upsell opportunities for new accounts or additional services. In the Fourth Industrial Revolution, analytics is no longer a trial-and-error science project — predictive and prescriptive analytics that are trusted and true should come standard.
- AI lets you have a conversation with your data. No longer do business users need to have an advanced degree in statistics or data analysis to interpret data. Questions like “What’s my open pipe for this quarter?” or “How does revenue for product A in 2018 compare to product B last year?” used to require coding and a deeper understanding of advanced analytics. Today’s natural language interfaces let you talk to your dashboards by typing or speaking questions to surface charts with answers instantly.
- AI lets machines handle the mundane. AI frees up your workers from mundane tasks by automating basic workflows, giving them the time to focus their efforts on the most important tasks that require human input and decision making. Once automated, simple actions like approving a discount under a certain amount, or escalating a complex service issue will be handled faster and managers have more time for more pressing issues.
AI we can trust
When it comes to predictive insights, it’s understandable that users are hesitant to trust new technologies that they don’t understand. After all, it wasn’t enough for your high school math teacher to write out the answer to an equation; they had to show us how they arrived at the result. The same is true for AI; we can only trust the data and recommendations if we know how the technology got there. The following capabilities are now a must-have for any AI powered analytics product:
- Transparency: Augmented analytics products should allow the user to have full access to the underlying model used to generate either the diagnostic story or prediction. The model should be exportable so that a customer can perform an independent validation.
- Explanation: The model needs to explain itself and the explanation needs to be in plain human speak. Just outputting a predicted number or score is not good enough. Along with it, the product needs to provide an explanation of the factors that drove that result — what factors can be attributed to a customer’s decreasing demand of a product? What factors should be taken into consideration when determining whether to pursue a lead?
- Accountability: The product should provide model performance measures for evaluation. Whether it’s GINI, R^2, max accuracy, MAE, or others, there are no shortage of success metrics. Augmented analytics should submit itself for a transparent evaluation and we should hold AI accountable. Additionally, a feedback system can tell which recommendations resulted in actions and if the desired impact was achieved. Finally, the product should also provide capabilities that guard against bias and allow designating certain fields as protected (e.g. race, gender and zip code, to name a few).
The future is here
I predict that within the next few years, artificial intelligence will be a standard feature of almost every modern Analytics platform. Just like airbags in a car today, AI will no longer be a shiny extra feature, but an integral component of any effective BI tool.
When general business users are able to take on tasks in a few clicks that used to require weeks of work by advanced data scientists, everyone benefits. Each employee is able to up-level their skills when there is technology to guide them. Your teams can spend more time engaging with customers in a meaningful way. The analytics industry has talked about actionable analytics for awhile, but AI is finally making the needs of the modern business user a reality.