Considerations in the Applicability of Decision Intelligence (DI) to both Strategic and Tactical Decisions
At the time of writing this post, we were in the midst of teaching Getting Started with Decision Intelligence, in our 9th cohort. A student asked today, “What’s the applicability of DI to different levels of decision making? Can it be used for both tactical and strategic decisions, or just strategic ones?”
The short answer is that DI is not just for complex strategic decisions—it can absolutely be used for tactical and operational ones as well.
For instance, Dr. Pratt and her team are currently funded by the US Department of Agriculture* to support sweet potato farming. One of the key decisions that they model is a daily (tactical) choice made in the packing plant: A packing manager has to decide the right bins to be used to fill a variety of sweet potato orders (e.g. from grocery stores and pet food companies). A packing manager has to choose the right potatoes to be used to fill each order, while minimizing the number of potatoes that cannot be assigned to any order. Structuring this choice as a CDD has been an important part of this project.
Here’s an example of a CDD from that project:
The bottom line: DI is particularly useful for strategic decisions because those are often the most complex, involve multiple stakeholders, and have long-term consequences.
However, tactical and operational decisions often benefit from the same clarity, structure, and alignment that DI brings, especially when decisions are recurring or part of a broader system. DI is not just about big, one-time strategic decisions—it’s about making better decisions at all levels by ensuring they are structured, data-informed, and aligned with desired outcomes.
tl;dr: there are complexities in decisions at both strategic and tactical levels that justify the use of DI.
As you think through whether DI is applicable in a particular situation or not, there are other factors to keep in mind, which transcend the tactical vs. strategic dimension. These include the following:
- Is it an action-to-outcome decision?
- Do we have target fields from historical training data (if so, then machine learning should be considered for the whole action-to-outcome flow)?
- Is the decision complex enough to justify modeling it?
A few side points:
- Strategic vs. tactical is not the same as manual vs. automated.
- One-off vs. repeated decisions is another distinction, which sometimes – but not always – corresponds to the need for manual vs. automated decisions.
To illustrate this, a strategic decision (e.g., “Should we expand into a new market?”) might be a complex one-time decision, while a tactical or operational decision (e.g., “How much inventory should we restock this week?”) might happen on a recurring basis. However, either of these could be handled manually or automated with the right data and decision models. Each combination of one-off vs. repeated, tactical vs. strategic, and automated vs. manual may require a different approach.
Another side point: For automating decisions, the results of CDD simulations can be captured in decision rules. See this related post for more on this extensional vs. intensional decision representation topic: a subtle but important distinction that is rarely understood in the DI community.
Finally, whether or not a CDD-based simulation is ultimately used, collaboratively creating and sharing the CDD drawing is still often highly valuable.
Interested in learning more? Check out the Decision Intelligence Handbook to learn more details about where DI is or is not the best choice.
*This research was supported in part by the intramural research program of the U.S. Department of Agriculture, National Institute of Food and Agriculture, Agriculture and Food Research Initiative (AFRI) Data Science for Food and Agriculture program, Grant 2022-67021-37137, Project Number NC.W-2021-11509
Liz Trotman
Liz is a Lead Data Analyst at Western Governors University. Liz has been with WGU for nearly 15 years and is an expert in student and faculty analytics. Prior to working in education, Liz worked in the Information Technology as a Network and Systems Administrator.
Computer scientist Dr. Lorien Pratt is one of the pioneers of artificial intelligence and is credited with inventing transfer learning. Following a 36-year career delivering applied AI systems, along with leadership and program committee responsibilities at the industry-leading NeurIPS conference, Pratt is recognized alongside Marie Curie and others by the Women Innovators and Inventors Project. Today, Pratt is chief scientist at www.quantellia.com, where she continues to push the boundaries of technology as a creator of and evangelist for Decision Intelligence (DI), a cofounder of www.opendi.org, and through building Large Language Models (LLMs) integrated with DI and ML. Pratt’s The Decision Intelligence Handbook (O’Reilly, 2023) is receiving worldwide attention. With recent publications in Foresight, Futures, and Frontiers, Pratt's DI innovations connect AI to human decision making in a way that is transparent, auditable, and accessible, bringing this important technology to the masses, and reducing information inequality.
