5 Things to Consider Before Applying AI to Analytics | MicroStrategy
BI Trends

5 Things to Consider Before Applying AI to Analytics

IDC predicts that by 2020, the total amount of digital data created worldwide will equal 44 zettabytes. By 2025, that will rise to 163 zettabytes, ballooned by the growing number of devices and sensors.

IDC also predicts that the amount of the global datasphere that’s subject to data analysis will grow by a factor of 50—and more than a quarter of all data created will be in real time. What’s more, 60% of those 163 zettabytes of data will be created and managed by enterprise organizations.

AI to augment insights and decision-making with speed, precision, and at scale will soon no longer be a nice-to-have; it will be a necessity. Constellation Research predicts that the artificial intelligence market will jump from $40.3 billion in 2020 to $100 billion by 2025 as enterprise organizations look to deliver on a spectrum of seven outcomes:

  • Perception (what’s happening now?)
  • Notification (what do I need to know now?)
  • Suggestion (what do you recommend?)
  • Automation (what should I always do?)
  • Prediction (what can I expect to happen?)
  • Prevention (what can I avoid?)
  • Situational Awareness (what do I need to do right now?)

When it comes to analytics, the key will be to make these insights understandable and actionable to even the non-technical business user—but more importantly, to be able to trust the AI-augmented results. In ensuring the latter, Constellation Research Principal Analyst, Founder and Chairman and Disrupting Digital Business author Ray Wang notes five essential elements of AI-augmented data systems:

  1. The systems must be transparent. You have to understand how the system gets to its decisions and why and how an algorithm is created so people know the implications. 
  2. It’s got to be explainable. When you understand the attribution, you understand the impact of what could happen next and why certain decisions or best next actions have been suggested.
  3. It’s got to be reversible. You need the ability to reverse decisions, so that if you learn something new, you don’t continue to do the old or the wrong thing.
  4. It’s got to be trainable. You have to pair people with systems and train the systems over time, so that you and the system know what’s going on. 
  5. It’s got to be human-led. The process must begin with a human and must end with a human. When you want to shut down a process, you can.

“These are important overall design points when you think about digital transformation,” says Wang. “They apply to AI and ethics. They apply to how we handle and work with data.”

To gain and retain the competitive advantage needed for survival in an era of Digital Darwinism, enterprise organizations will need to quickly (yet astutely) embrace artificial intelligence. Learn more about this and other trending topics from Global 2000 adviser Ray Wang in the 18-page Q & A-style ebook, The Role of Data in Digital Transformation.

Comments Blog post currently doesn't have any comments.
Security code