December 21, 2021
Defining Decision Intelligence
We sat down with Dr. Lorien Pratt, a pioneer in the field of decision intelligence, to discuss the market momentum and adoption of decision intelligence software and frameworks. Dr. Pratt is also the co-founder and Chief Scientist of Mt. View-based Quantellia, which offers data, analytics and decision intelligence software and services worldwide.
Decision intelligence (DI) is a practical discipline that is used to improve decision making by explicitly understanding and engineering how decisions are made and how outcomes are evaluated and improved by feedback. DI bridges the gap between AI (which predicts and labels) and humans who, in many situations, think in terms of making decisions about actions that lead to specific outcomes. It answers contextualized questions like, “If I take action X, in situation (context) Y, what will the outcome be for my business or in relation to a certain KPI.”
Decision intelligence (DI) is a great example of how we can use AI in the business context to improve decision making and, in turn, improve outcomes for an organization. A decision intelligence framework helps organizations with operationalizing AI to augment human decision making, in order for humans to more confidently and accurately decide on which actions to take in order to drive better outcomes. From a practical perspective, this means bringing together human knowledge with the data, analytics, and machine learning capabilities needed to create a system that can recommend the best action to achieve an outcome and then ensure it continues to provide the best action over time as the situational context changes.
I work with a lot of data teams in government, business, nonprofits and other organizations. They are unanimously frustrated by the spreadsheets, reports and dashboards that data scientists are delivering. Specifically, despite an increasing amount of data and cheap computational power, they find that they still can’t make better decisions. What they are starting to realize is that their customers (decision-makers inside their organizations) don’t want simply insights or pretty visualizations – what they really want to know is if they take an action based on an analysis presented by the data scientist, what are the implications, risks, and financial impacts? How does that ultimately impact the things I care about or am measured on at work? This is the gap that DI fills: it is the “last mile” for making AI useful to humans and to effectively support their decision -making.
Another thought leader in the field, Cassie Kozyrkov, Chief Decision Scientist at Google, also describes decision intelligence as a way to help humans take better actions that lead to improved outcomes. She advocates that a combination of multiple disciplines including data science, neuroscience, psychology, economics, decision theory, and managerial science must be integrated to make better decisions. And I agree wholeheartedly. Cassie’s role at Google is an example of the type of new leader that data-driven organizations are starting to weave into the business fabric in order to make the most out of the treasure trove of data and human knowledge that they have available, and empowering people to go beyond just insights to actions and business outcomes.
Decision intelligence is the layer in between the technology and humans who ultimately make the decision. As shown below, DI requires a “human in the loop.”
Decision intelligence connects AI and human decision making. This means that, rather than a decision made solely by a human or a decision made by a computer, it is the best of both worlds. And it is now taking off. DI is on the Gartner hype cycle and listed as a top trend for 2022, and there are myriad decision intelligence software companies getting funded – this also signals that it is a maturing discipline that organizations should be paying attention to and planning for.
Q – What is different about AI, BI and DI?
I began evangelizing what would now be known as decision intelligence in 2009. I was pretty alone in this space for a few years, but I’m happy that, in more recent years, Cassie Kozyrkov from Google, research firm Gartner, and other thought leaders picked up on it. Unfortunately, we’ve also seen companies that are using the name “decision intelligence” in a way that’s inconsistent to the rest of us: simply as a new label for AI, in part to help to escape some of the reputational problems that AI is beginning to suffer from, as we see more and more AI failures.
DI was built to bridge some gaps that were left by AI and BI. BI is the idea of ensuring that humans have easy access to the most up-to-date information, in a way that is as easy to use as possible. AI is a bit fuzzier, as its definition has changed over the years. But most of today’s successful AI either labels things (“this is a picture of a cat”), predicts things (“this security will increase in value in the next 12 months”), or does natural language processing, as we see in, say, Google Translate.
But humans want answers to how to achieve business outcomes, neither of which is captured above. AI provides “insights”, BI gives us data, but there is something missing: a connection of those insights and into the core question that I learned in my market research that people want answered that I mentioned above, “If I take this action, in this context, what will be the business outcome?” A related question is “what is the best action that I can take, to achieve the best outcomes, in this context?”
Neither of those questions are fully answered by AI or BI. But we can use AI and BI in combination with models from other disciplines to answer those questions. DI is that “glue” layer that connects humans to AI, BI, and other disciplines, and also helps teams of humans (even without technology) to get on the same page. I talk about how to do this in my book: Link: How Decision Intelligence Connects Data, Actions, and Outcomes for a Better World (www.linkthebook.com), and I also have a second book that will be published in 2022 which is a practical handbook providing a step-by-step process for making this connection.
Q – Will DI replace BI in the enterprise?
Not at all. DI needs BI, and vice versa.
Today, more data is available than is currently used for meaningful analysis. In many situations, is no longer an issue of not enough data. Companies want to go beyond insights (and fancy presentation of those insights) to know what will happen if they take actions and how those actions will impact their business. Human insight + AI are required to take advantage of the troves of data organizations hold and turn it into effective insight for active decision-making.
The last mile – technology to outcomes – tends to be misunderstood in the marketplace. Business intelligence augments DI, but it’s not the same, much like AI is not the same as DI.
BI can be a massive adjunct to decision intelligence and shouldn’t be ignored. BI systems and strategies are being augmented with AI and machine learning to provide decision-making context and recommendations across the enterprise. This makes them better at telling us what can be concluded from the data. BI is still not able to tell us what decision needs to be made – this is the “last mile of analytics” problem. But BI is great at tracking the assumptions underlying a decision and alerting us to when they change enough that the decision must be revisited.
Q – What are the main drivers for DI?
Both good and bad factors are driving DI. The bad news is, AI is in the process of entering its third or fourth “AI winter”, so some marketers are trying to sneak it in as DI, saying “here’s our data, here’s our AI, and these can be used to support your decisions” without substantively addressing the question of how to help their customers connect action-to-outcome decision making to these technologies. Another error is that companies are still focusing too much on fully autonomous decisions (those without a human in the loop) – that set of uses cases is pretty much tapped out. The great value of AI in the next 10 years will come with AI working hand-in-hand with human experts in scenarios that Gartner calls “Decision Augmentation” , which is distinct from full automation. (I like to call it “Intelligence Amplification”, in homage to Douglas Englebart who first described this idea in the 1960s).
But the good news is, BI and AI failures are leaving this last mile on the table. Companies often have a very expensive analytics tech stack. They’re required to show ROI on these stacks, and it’s hard to do without that last mile. DI closes this gap, as has been demonstrated by a number of projects over the years with government organizations, very large financial institutions, nonprofits, and more.
Something that’s often overlooked with DI is the human element. Change management and making sure you’re using the right solution for the problem are critical to the success of a DI initiative. And organizations are starting to make sure they’re set up with leadership involved: chief decision officers, decision architects, decision engineers to take the architecture and build model specifications.
All of this is a precursor to what will make it much easier for decision makers to understand what is the most valuable data for a decision, and to make informed decisions based on the most up-to-date understanding of their context.
Decision intelligence is something that every business leader should understand, especially if their company has a “data-driven organization” mandate or has made a big investment in data infrastructure. While we cannot make every C-level executive an expert in classification methods or linear programming, we can make them aware of how data, AI, and human knowledge can work together to help them achieve optimal decisions. You can read about this all in my most recent book, at www.linkthebook.com .