the decision intelligence framework is perfect for guiding AI to consistently produce business value for you, the methodology can be used with no digital data or machines. For example, you can use AI to make decisions on a spreadsheet, on the back of a napkin, on a single sheet of paper, or even on a wall (using a crayon, of course). That’s because the process you use is up to the decision-maker to choose. The Decision Intelligence process itself can be quick and short, or it can be quite complex and take some time to complete. You may want to start with a SWOT table listing the Strengths, Weaknesses, Opportunities, and Threats when making your initial decision. From there, you can determine the steps you need to take to make your decision render a desired impact in the real world.
The process is similar to determining a destination and then mapping out the best route between where you are and where you want to be. It’s the impact you desire, however, that will define which route is best. Need to be there fast? The direct route is best. Want to see more along the way or stop at tourist attractions? Then a scenic route is the best way. Want to use your hotel rewards points or your gas rewards card on the trip? Then mapping a route based on the location of certain hotel and gas brands is the best route.
Working within a decision intelligence framework forces you to become more aware of how the decision-making process works. For example, many of the mental processes you use are intuitive — that's what makes it possible for you to come to conclusions quickly. But make no mistake: Whether you realize it or not, your brain is calculating the same mathematical formulas as a machine would use to help you reach the same conclusions. There’s a simple reason for that: Machines copy how people think. As such, machines are definitely the sidekicks in decision intelligence processes, there to assist and augment your efforts.
Tracking the Inverted V
You may be wondering how the processes in decision intelligence differ from those used in data analytics. After all, it’s obvious that decisions are also made first when using data analytics in the usual way. For example, someone decides what the business rules are before they apply them to data analytics or AI. Someone also decides what data to use, what data sources to join, and what queries to make. Further, someone decides what projects to launch and whether to send them to production. And so on.
With all these decisions upfront, what does “Put the decision first” in decision intelligence mean? And how does it change anything? It helps to remember that the process in machine-based decision-making is linear, meaning that it moves consecutively from data preparation and selection to algorithm inputs and, finally, to an output. The output is typically an insight or a recommendation delivered as a visualization, as narrative text, or as both, from which a human can decide what action to take. Sometimes, the output is connected to an automated process that then takes an action as directed by the output.
In any case, the path is a straight line.
Now tilt that line upward so that it’s the first leg of an inverted V. At the bottom is the starting point, which is the data to be analyzed. At the top is the decision to be made based on the analyses. That’s your path upward.
Ignore that path and work your way back down from the decision to the data. Rarely will you follow the same path down. Instead, you’ll create a different path that will be more specifically tied to the decision. The two paths together resemble an inverted V.
The first leg of the inverted V begins with mining the data, and then an analysis follows. If you think about it, this process is now defining the decisions you can make. By contrast, in the V's second leg, the decision is defining the data, tools, and queries.
The first leg is a discovery mission. The second is a mission with a purpose.
Which leg do you think will consistently deliver a payload?
And that, my friend, is why and how you put the decision first.
Estimating How Much Decision Intelligence Will Cost You
Ah, yes, the bottom-line question on everyone’s lips is cost. Certainly, cost is a major consideration in nearly every business decision. This time, however, it isn’t much of an issue. Because decision intelligence is a rethink and not a redo, you likely already have in place many of the technologies and tools you need. (Think of it as leveraging those items to produce a higher return on investment, or ROI, on what you already have.) Of the tools you may not have, many of the products you need offer free versions or at least free trials so that you can see how they work and whether they’re a fit for your organization.
That may leave a few tools to buy, depending on your current mix of technologies. All told, it’s rarely a huge expense to switch from data mining tactics to decision intelligence.
The following checklist can help you form an idea of some of the technologies commonly used in decision intelligence. That way, you can quickly see what you may need to put on your shopping list — or which functions you might want to hire a third-party who has these things and the experience to use them to do some of this for you.
Decision modeling software is a part of decision management systems that represents business decisions via a standardized notation — often the Business Process Model and Notation (BPMN) standard — that is used by business analysts and subject matter experts (SMEs), rather than developers. Examples include ACTICO Platform, Red Hat Decision Manager, and FICO Decision Management Platform.
Business rule management software to manage the business rules in decision-making. Sometimes these are standalone software products and sometimes they are part of a decision management system as well. Examples include VisiRule, Red Hat Decision Manager, SAS Business Rules Manager, InRule, and DecisionRules.
An AutoML stack or another collection of software capable of automating all or part of the building of ML models. AutoML simplifies the machine learning model developer process by automating many of the more laborious steps, such as feature engineering, hyperparameter optimization, and creating the layers in the neural architecture. Don’t worry if you don’t quite grasp what these automated activities entail because the point in having AutoML is to do all that complicated and time intensive stuff for you. The cool thing is that while AutoML is a useful tool for data scientists, it’s just as useful in democratizing AI. Yes, you too will one day make the AI you want to use in your DI process — by telling AI to make it for you. See, not as hard a concept as you thought. Examples of AutoML vendors include DataRobot, H2O.ai, and Google Cloud AutoML.
A good data platform which is a technology that bundles several big data applications and tools in a single package. Preferably get one that supports both the creation of algorithms and the delivery of transactional data in real-time. Examples include Google Cloud AI platform, RStudio, TensorFlow, and Microsoft Azure.
A BI app with natural language processing,
