AI-Infused BI: Only as Good as Its Transparency
AI will wend its way into BI, making it better, but there's a hitch.
- By James E. Powell
- October 10, 2018
Business intelligence is entering a new era. AI is becoming a mainstream part of BI technologies and practices and will eventually find its way into all aspects of BI -- from discovery and data collection to analysis and visualization. Analytics will embrace AI techniques and technologies to increase its speed, depth, and sophistication; as a result, enterprises can make better decisions faster.
There's a hitch, however.
In the recent TDWI Checklist Report: AI for BI: Six Strategies for Augmenting Business Intelligence with AI and Machine Learning, David Stodder, senior director of TDWI Research for business intelligence, notes that AI's usefulness hinges on whether users can trust the insights AI-enabled BI provides. Are the results accurate, relevant, and transparent?
There's no doubt AI is both powerful and increasingly ubiquitous. Stodder notes that "AI use cases stretch from recommendation engines to capabilities embedded in applications and devices for self-driving cars, medical devices for detecting cancer, and voice-controlled personal assistants." Thanks to the power of [massively parallel processing] systems and techniques including machine learning, deep learning, and natural language processing, organizations can "discover patterns and other insights hidden in sometimes billions of data points with less human intervention and greater speed and efficiency."
Stodder's report explains how enterprises can adopt AI-enabled BI to "accomplish projects faster and provide relevant and accurate insights that users can trust." First on his list: AI can enhance the speed, number, and quality of decisions. AI isn't just about automating functions; it's about enhancing intelligence so humans can make better decisions.
Another benefit: AI can provide a broader view of both traditional and new data sources. In fact, the more data AI enables users to work with, the more efficient they will be with data discovery. Users will also gain the ability to ask new types of questions and explore new perspectives -- without IT's help. "The augmented intelligence provided by AI can enhance and expand self-service capabilities further to let nontechnical users engage in more relevant data interaction," the author explains.
Trust Is Key
Users will need to understand AI technology as more than just a "black box." Accuracy and consistency have always been core requirements for BI functions such as performance management and operational reports -- that doesn't change when AI is added to the toolset.
Trust becomes even more important as self-service analytics encourages users to explore, test, and examine different variables to develop new perspectives. The problem as Stodder sees it is that "many AI techniques vary the path to accurate insights [using techniques such as] statistical and mathematical analysis that produce results that must be interpreted, not just accepted." [Emphasis added.]
He warns that organizations that infuse BI with AI must prepare decision makers to "understand how the insights were derived, learn how to interpret results, and ask more questions to refine their insights. It is through such processes that users build trust in AI and analytics."
To be effective, he recommends every organization make explaining results part of its routine. Just as the GDPR has introduced the idea that consumers have the "right to an explanation" for how certain data-driven decisions are made, BI users will expect to understand how an AI system came to a conclusion.
Stodder says TDWI expects to see technology and practices for increasing transparency becoming a core capability in AI systems in the future.
You can read the full report here. Visitors new to TDWI must complete a short, one-time registration for access.
James E. Powell is the editorial director of TDWI, including the Business Intelligence Journal and Upside newsletter.