top of page
Contact Us

Understanding Evolution of Data Before Implementing AI

This is a year of articles and conferences about artificial intelligence and machine learning (ML). Even though ML has made strides in the consumer world, it has only started to show value in enterprises. Many companies are still trying to figure out what value they can harness, how much data they have, how much data is enough, how to start or pick a project, how to measure return on investment (ROI) and how to use ML as a tool in their digital transformations. I plan to write articles that address each of these areas.

Evolution of Data Before Implementing AI
Evolution of Data Before Implementing AI

It is important to understand the evolution of data within the enterprise in order to understand the true value of machine learning. When the internet started, the focus was on transaction response time within online transaction processing (OLTP) systems. These systems powered websites, and the focus was on improving reliability, availability, and scalability (RAS). As transactions grew on e-commerce sites, we wanted to know who was using the system, from where, what they were doing, for how long, and most importantly how we could offer more value to existing customers and bring in new customers. This brought a whole slew of applications around business intelligence and reporting or analytics engines.

We went into a world of user-generated content -- both structured and unstructured through social media and smart devices including mobile, voice, and video content. To garner value from this content, new types of analytics engines were needed. The growth in computing power along with a lot of research to create sophisticated algorithmic tools has allowed for the ability to fundamentally change what a platform is all about.

Traditionally, a platform was used to address an enterprise process workflow -- human resources (HR), finance, manufacturing, etc. They are what we categorize as enterprise resource planning (ERP), customer relationship management (CRM), human capital management (HCM), functional setup manager (FSM), information technology operations (ITOps), etc. The data generated by these workflows were then analyzed using analytics or business intelligence applications to make further modifications to workflow. These workflow applications were customized as the data warranted any changes in the workflow.

When intelligence becomes the new platform, data from these traditional applications will be used to determine the workflow and actions of organizations. The workflow actions will be passed on to the traditional applications or directly to the people or system that will perform the actions. These new systems of intelligence will emerge and will force existing workflow applications to change to be end-user targeted. We are already seeing a trend where AI platforms are slowly becoming a playground for new intelligent applications. More importantly, because open-source intelligent platforms in this area are as rich as enterprise platforms, we are also noticing new generations of applications. These applications prescribe specific actions that can be taken in their field of expertise -- HR recruitment, personality analysis, service optimization, sales upsell, etc.

Take commercial travel as an example. Early websites pointed out the lowest flight price available from point A to point B. Then came websites that compared other websites and aggregated results. Now, we have the ability to choose a budget and have a website or app suggest destinations that fit that budget or better days for us to travel. Our decision has transformed from just getting data and then taking action to an action being recommended to us that we can then decide whether or not we want to pursue. This works best when we have clearly established our goals (in this example it is the budget).

This same level of intelligence has not come into enterprise applications. In a way, they are lagging behind. From what we have seen, it is not a lack of data causing this lag, but issues with culture and defining our goals.

I will discuss more how enterprises can choose what use cases are best for implementing AI, how much data is good enough when to build instead of buy, and how to measure return on investment in subsequent articles.

Learn more,

bottom of page