Business intelligence has evolved. It now includes a spectrum of analytics, one of the most promising of which has been described as augmented intelligence. Some organizations have used the term to describe the practical reality that artificial intelligence with machine learning is not replacing human intelligence, but augmenting it. The term also represents the application of AI/ML to make business intelligence and analytics tools more powerful and easier to use. It’s this latter usage that I prefer and I’d like to explore in this perspective.
I see three main types of augmented intelligence:
Automated AI/ML analyses are becoming more common among analytics vendors. Many AI/ML vendors tout these capabilities, and some have built businesses around automated or assisted AI/ML model development. While these capabilities are valuable, I remain a skeptic as to whether an organization could completely automate the development and deployment of models without any further oversight or governance by highly trained AI/ML specialists. We’re getting closer, but I would still want a review of any models developed automatically before they were deployed into production.
However, automated “insights” can be quite valuable. BI vendors are now running automated AI/ML analyses to help guide individuals as they examine the data in their organizations. No special expertise is required. The analyses offer an assessment of which metrics may be most significant or which factors are causing the largest changes in performance. In addition to performing AI/ML-based analyses that many individuals wouldn’t have the skills to perform, these automated analyses also introduce some consistency across the organization. The same type of analyses will be conducted and available to all.
Natural language processing also employs AI/ML on behalf of individuals using the software. AI/ML is used to interpret queries and translate them into SQL or other computer languages as necessary to retrieve the information for which an individual is searching. Similarly, AI/ML is used as part of the natural language generation process to produce sentences or paragraphs of text describing what was found in the data. Many people do not know how to interpret a table of numbers or a chart of data. NLG helps draw their attention to critical pieces of information driving the organization’s performance. Again, this technique introduces more consistency in the organization, rather than leaving to chance how a set of data might be interpreted. Organizations using NLP are more likely to report having more than half the workforce using analytics (31%) compared to those relying only on reports, dashboard and visualizations (18%).
Finally, AI/ML can be used to make data and analytics software products easier to use. Many of the automated insights referenced above produce visualizations as part of the analyses. Those visualizations can form the basis of a new dashboard. AI/ML can make suggestions on what data to analyze. For instance, an individual’s department or location might be used to suggest the filtering to be applied to data. More sophisticated assistance might include considering what others in a similar role are analyzing. Many analyses require data to be joined together, and AI/ML can assist with that process as well.
Regards,
David Menninger