Why “Data Intelligence” is not quite the same as “Decision Intelligence”?
Information travels fast and far than ever before, causing us all to fall victim to the ‘meme’ of the day. Big data, ML, and AI are examples of the recent memes in the digital enterprise context. Somewhat more recent is the ‘Decision Intelligence’ meme. Quite a few technology vendors have started positing themselves as ‘Decision Intelligence’ vendors regardless of whether or not they have anything to do with decision intelligence. Such vendors contrast their offerings with traditional BI offerings, and as they readily differ from such offerings, they automatically park themselves in the decision intelligence camp.
I will use the term ‘Data Intelligence’ to describe products that are beyond BI in some way but do not qualify as decision intelligence products. As the discussion concludes, I propose a set of questions for vendors to discern their offering from a decision intelligence perspective.
What is “Data Intelligence”?
Data resides in a data repository. It may have been created and uploaded by a human agent or automatically captured and stored by an automated agent. A software system that allows human agents to comprehend data in some way provides data intelligence.
The comprehension of data goes beyond simple querying of content. It includes all that the software system can infer from this data alone about the past, present, and future about the domain of data. One can search for the presence or absence of certain elements in the data, perform sorts and aggregations on data elements, and find co-relations among different parts of data. Furthermore, if the data is labeled, one can also predict the future state based on past patterns embedded in the data. The so-called descriptive, diagnostics and, predictive analytics generate data intelligence artifacts that can be produced on-demand and presented to a human agent when requested.
Though a valuable facet of decision intelligence, data comprehension is not decision intelligence. This point becomes apparent by realizing that decision intelligence support is possible even without preliminary data.
Snapping a natural language interface on a “Data Intelligence” system does not make it a “Decision Intelligence” system.
We have evolved from using explicit SQL queries, DSL queries, and keyword-based searches to full-powered natural language conversations to feed our requests to a data intelligence system. Advanced modalities of human-machine interaction indeed make data intelligence highly convenient and widely accessible to human agents. Though easy to access, the purpose served is still data comprehension and nothing more.
Snapping a fancy data visualization interface on a “Data Intelligence” system does not make it a “Decision Intelligence” system.
A properly conceptualized data visualization, especially interactive visualization, has the potential to accelerate the speed and depth of data comprehension. On the other hand, a poorly designed data visualization may also lead a human agent astray.
Snapping prediction engines using ML/AI models on a “Data Intelligence” system does not make it a “Decision Intelligence” system.
Adding prediction engines to a data intelligence system enhances data comprehension as it surfaces hidden patterns in the data that are otherwise invisible to descriptive analytics methods. As predictions provide a glimpse into the future, it is not a stretch for someone to confuse it with decision intelligence.
So, what is Decision Intelligence?
As data intelligence pertains to data comprehension, decision intelligence is about comprehending the process, methods, and the human side of decision-making. Decision Intelligence starts with explicit modeling of decision networks (not data) to adequately frame a decision-making problem, helps human agents to understand their choices by enumerating and evaluating them from a value and risk perspective. A decision is rarely in isolation. Prior decisions influence a new one. And current decisions are mindful of future decisions to accomplish an overall objective. Furthermore, the description of the decision-making scenario may be fraught with uncertainty and often change as decision-making activity progresses.
“Data Intelligence” is neither necessary nor sufficient for “Decision Intelligence”.
Human agents can make decisions even when there is no data available. As data becomes available, they can incorporate data intelligence into the mix. A decision intelligence system can assist human agents inappropriately framing the decision-making problem, help choose problem-solving methods, and automate the orchestration of the decision-making process among multiple stakeholders, often with conflicting objectives.
Data Intelligence can be fully automated whereas ‘Decision Intelligence’ involves man-machine collaboration.
As data is formally and fully described, a data intelligence system is fully automatable to provide rich support for data comprehension. On the other hand, decisions are rarely made by machines alone, except in narrow and fully defined settings such as tasks by autonomous robots in an assembly line.
Providing decision intelligence support requires an understanding of how decisions are made in a decision-making context
The software system for decision intelligence support must understand how human agents collaborate and negotiate in general and in particular situations. How they perceive value, and what is their tolerance for risk?
A decision intelligence system can propose to delay a decision.
A decision intelligence system does not necessarily recommend a decision choice. It may propose delaying a decision until there is more clarity about the situation and further propose actions to enhance situation clarity.
Few Questions for ‘Decision Intelligence” vendors
One can pose a few questions to decision intelligence product vendors to discern their coverage and depth of decision intelligence.
- Does the product explicitly model and maintain the decision network?
- Does the product provide any assistance in framing the decision-making problem?
- How does the product handle information uncertainty?
- Does the product allow change management of decision problem attributes?
- Does the product support conflicting objectives?
- Does the product support tradeoff and negotiation among multiple stakeholders?
