Understanding the difference between Business Intelligence (BI) and Artificial Intelligence (AI)

by Kumar Singh, Research Director, Automation & Analytics, SAPinsider

A Brief history of Busines Intelligence (BI) systems

The world of analytics owes a lot to Edgar Codd’s paper “A relational model for data for large shared data banks”. Published in 1970, it heralded the world into next generation relational data bases. This new generation of databases allowed for a much larger capacity to store and manipulate data. IBM initially was reluctant to implement Codd’s design since it was making significant revenue from its existing data base systems and that allowed other organizations to leverage that design.It became the predominant form of database and helped build the first version of the Business Intelligence (BI) tools and BI Industry.

Towards the end of 1970s, Oracle, the first true relationl database management system, was launched. And this was when Business Intelligence tools became much more powerful. Fast forward to the 90s, as database technologies became more and more accessible, BI tools flourished, leading to sandardized methodologies like ELT and OLAP, which are still extremely crucial to implementing Businesss Intelligence capabilities. By the year 2000, BI tools became a must have for majority of businesses.

Advances in computing and proliferation of world wide web made BI tools much more powerful and useful by mid 2000s. This was the time around which Google analytic was launched, Big Data was coined as a term and initial seeds of cloud computing got planted.With huge proliferation in the amount of data being generated, rapid avances in computing and the data storage devices becoming nimbler and more sophisticated, BI tools evolved into the form we see them now.

Advent of Artificial Intelligence (AI) and Machine Learning (ML)

In the world of technology, it will be an impossible task now to find someone who is not aware of the term data science. With the emergence of data science, Artificial Intelligence and Machine Learning tools appeared on the horizon. The world was introduced to the significant impact AI and ML tools could make to the way businesses plan and operate. Among all this hype, BI, which was the poster child of the world of analytics till then, started to lose its glamour. Some quick highlights of AI and ML tools are:

• Algorithms driven decision making and cognitive automation that can be leveraged on all forms of data
• Leverages the same data source (ideally) in production
• Algorithms “learn” continuously from the data, thereby reacting to changes in underlying data that is fed to them
• Primarily used for prescriptive analytics, predictive analytics, reinforcement learning, computer vision etc.
• Is generally more computationally intensive than BI tools
• An organization that becomes best in class with these tools would have mastered BI tools as well since analytics success is defined more by culture vs technology

A “Cliff notes” comparison of BI and AI

Before we jump into the comparison, let us re-visit BI again in “Cliff Notes” format:

• Primarily descriptive analytics
• Provides insights on how the business ran or is running
• Can be analytics on historical data or on near real time data
• Some tools may have prescriptive analytics features as well
• Even though aspects like augmented analytics has made BI more self service, it still leverages manual touchpoints significantly
• Helps build data driven decision making culture
• Lays the foundation for more advanced forms of analytics
• When we talk about IoT enabled visibility (only), that also is a form of business intelligence
• Will always be the foundation of analytics in organizations though both BI and AI will merge in same tool sets in the future

As you can see from the points above, BI tools lay the foundation of analytics in your organization. And no matter how much AI and ML you leverage, BI will always be leveraged in some form though I expect that in the future, they will eventually get embedded in enterprise AI tools. The illustration below revisits AI and BI capabilities again and also illustrate how AI is actually not competing against BI but an extension of any organization’s analytics journey.

So here are some key points of comparison, as evident from the example in the illustration above:

• While descriptive analytics based business intelligence provides you all the “intelligence”, there are generally no prescriptions, predictions  or recommendations made. AI tools generally have the capability to “learn” from the data as well and hence generate prescriptive actions, predictions or recommendations
• Though AI systems and algorithms may leverage the same data source as BI tools, they are generally more data intensive. This is because to help them “learn”, they need to see a significantly large number of data points
• They are also computationally more intense than BI tools. To help an AI algorithm “learn”, it is “trained” on a large amount of data. “training” in simple terms is showing the algorithm data points with results already labeled (ex: Features of an image with the label that the image is that of a “Cat”). This training aspect is what makes these algorithms computationally more intensive
• Infrastructure wise, the amount of data being consumed and the computation requirement, both are significanly higher. You will hence see more and more AI platforms and solutions being cloud native or migrating to the cloud
• Though AutoML is helping develop more and more citizen data scientists, the skillset requirement to leverage AI tools is significantly higher than BI tools in their existing form

What does this mean for SAPinsiders ?

Amidst all the hype about AI and ML, there are certain key points that SAPinsiders need to be cognizant about:

• Though you will see journey maps, where descriptive analytics is one of the initial capabilities, it does not indicate that BI tools are inferior to other type of tools and algorithms. The fact is, an organization strategically needs to define what type of analytics it will be leveraging across its functions and processes. Eventually, you will be leveraging a portfolio of tools. And for some aspects, descriptive analytics is all that you need and anything more than that may be an overkill.
• When you think about embarking on building advanced capabilities with AI and ML, think about what capabilities you want to have in the future vs what can be done today. The fact is, any advanced AI or ML based solution will take time and patience to build. Trying to leverage them for operational analytics that you already execute successfully may not be prudent, though there are exceptions
• The only and the only true measure of having built advanced analytics capability is the numb er of citizen data scientists in your business processes. This is the most challenging capability to build since this requires much more than merely hiring a team of data scientists who churn out algorithms.