Technology is a powerful ally when it comes to improving decision-making processes. The key is to define which decisions should be optimized through digitization based on their impact on the business and their optimization potential. What questions do we need to ask ourselves to focus on the right decisions and maximize their value? What are the steps to follow in the decision optimization process?

The question is no longer whether machines can make decisions but what decisions we want them to make. The first question has been answered many times. For example, I could cite a historical date: May 11, 1997. That day, the world chess champion, Garry Kasparov, was defeated for the first time in his life by the Deep Blue program, created by IBM, in just nineteen plays.

The defeat of the Russian chess player made clear not only the ability of artificial intelligence to make decisions but its power of precision: the logic of machines is less likely to fall into specific errors that condition human judgment.

During the last decades, much has been written about biases that lead to systematic and verifiable errors in the results of a decision. In 2021 the Nobel laureate and expert in behavioral economics, Daniel Kahneman, published a book in which he introduces the concept of ‘noise’, as a reference to the variability of people’s decisions, subject to factors as unpredictable as mood or climate.

None of this affects machines, which is why they can decide with greater precision. The business world has taken due note of this advantage because it knows very well that a correct decision, made at the right time, can define the success or failure of a business. Hence, the digitalization of decisions is a crucial pillar of the incredible transformation that organizations are undergoing. We can think of it as a long process in which companies have been gradually incorporating different tools with which they have been able to evolve from analogical decision-making, based on unstructured information and intuition, to an almost totally digital process, in which algorithms or artificial neural networks can recommend alternatives and steps to follow. In the most extreme case, artificial intelligence can be configured to make decisions directly without human intervention.

This process has accelerated in recent years, having overcome some important barriers such as the low processing capacity of machines, the scarce and insufficient quality of information and data to implement artificial intelligence on a scale, or the natural resistance on the part of clients and industry management agents to abandon less systematic decision-making processes. Since then, the appearance of Big Data and Cloud Computing has multiplied access to information, while Machine Learning and Deep Learning techniques have strengthened their operation, two decisive factors for digitizing decision-making.

Identifying decisions that matter

It is becoming increasingly evident that a company that decides better has better results. If, in addition, machine intelligence can improve the decision-making process, what follows is to ask: what are the decisions with the most significant potential for digitization and positive impact on the results of a company? In general terms, we can think of a process that includes the following steps:

1. Listing all impactful decisions and prioritizing those with the highest expected benefit

The first step is to list all the decisions that significantly impact the organization’s results and choose which are worth considering for digitization. Starting by listing the decisions prevents us from making a typical mistake: processing and analyzing data without knowing why we are doing so and without clearly identifying the business objective behind it.

From the previous list, we will determine how each decision impacts the business. The aim is to distinguish those that have the most significant impact, either individually (a decision that by itself has a high impact) or in an aggregate way (a decision that is made many times and its recurrence through repetition yields an important value).

• Ask yourself: What decisions are behind the core business or area challenges? What decisions are made at each stage of the process to be evaluated? What is the aggregate impact of each decision on business results? How does the decision situation vary each time it has to be made? Are there errors or noise that could be reduced?

2. Understanding the nature of each prioritized decision comprehensively and defining the best solution to optimize it

This stage consists of checking if they are decisions for which we have (or can have) information and if we are able to structure them or whether they are decisions that have different alternatives or are subject to different risks each time they have to be made. Understanding this will allow us to distinguish which type of solution will be more convenient: whether we can reach full automation, whether it will be a development that assists the human responsible for making the decision, etc.

In this case, it is helpful to dwell on each of the components of the decision and assess the extent to which it is possible to create a repeatable approach. Likewise, it is essential to put in place initiatives to improve data quality and improve access to internal or external data sources.

• Ask yourself: Can the decision be structured with objectives and predefined business rules? Are there constant alternatives and variables over time? What information is needed to make the decision? Do we have (or can we have) the sources of information we need?

3. Comprehending the optimization potential to define economic feasibility

Once the decisions have been prioritized and the potential solutions have been identified, it is necessary to calculate their current cost in time and resources. Some decisions are very expensive to make and are precisely the ones that are most convenient to digitize. Likewise, we must understand what the cost of strengthening them with technology would be.

To evaluate the investment in a digitization process, it is also convenient to define how reengineering decision-making could move the organization forward; for example, if we plan to drive digital transformation or provide a competitive advantage.

• Ask yourself: What is the estimated value of the decision without decision intelligence? What are the costs in time and resources of making the decision today? What is the margin of error they have today? How much would optimization change the results for the business? What is the cost of incorporating technology, data science, and the necessary information to optimize them?

4. Building a technical solution prototype and defining the required organizational changes to make the solution sustainable

To implement the technical development that will allow us to optimize the decision, we must first understand the tool’s detailed functional requirements (what it has to do, what type of analysis, and what the output should be). This will enable us to develop later a simple prototype to test and adjust the tool while obtaining better results for the environment in which it is being tested.

It is vital that during this stage, we also set in motion the organizational changes that will be required for the solution to work and those that will have to be made as a consequence of the change. That is to say; there will be some that will be a necessary condition (e.g., guaranteeing the decision-making and analytical competencies required according to each profile) and others that will be the result of optimization (e.g., fewer routines, more straightforward and cleaner processes, changes in the governance model).

• Ask yourself: What calculation logic should the model have to optimize the decision? What requirements must a prototype meet to be launched as a pilot in production? To test it in different markets/impact areas, what adaptations must be made to the prototype? What adjustments need to be made in the organizational design, processes, roles, and responsibilities for digitization to work? What internal skills need to be developed for each type of profile to make the most of the solution?

5. Implementing, learning and adjusting

Once the design is complete, we will move forward with implementing the technical solution and the organizational changes. This is a highly variable stage, depending on the scope of the implementation and the impact it will have on the organization.

In any case, the key is to keep the process alive to ensure continuous learning on the technical side (identifying adjustments to be made to the tool) and on the organizational side (evaluating additional changes required for the solution to work).

• Ask yourself: What is the ideal implementation methodology (modules, geography, channel, etc.)? Does the developed model cover the needs of different markets? How can we improve the model’s accuracy through information or new correlations? How do we ensure continuous feedback to improve the model? Are the organizational adjustments adequate to optimize the solution?

On a day-to-day basis, it is possible to see these implementations in sectors such as Consumer Packaged Goods, which is carrying out a great digitization process of its marketing and commercial process. For example, companies in this sector have many salespeople who visit points of sale daily and constantly make decisions to apply discounts or other commercial conditions. Although these decisions have a limited number of alternatives, within certain benefits that can be granted to each client, sellers often make the final decision, and there is much variability when comparing the chosen option to similar situations. It is true that the decision of a single seller has a low impact but think of that small impact multiplied by the number of sellers making that same decision across geographies. The aggregate impact changes quite a bit, doesn’t it? For this reason, many leading companies are automating these decisions so that, given specific customer characteristics, the recommendation of the trading condition to apply is through business rules that enhance profit maximization.

The digitization of decisions is only part of the potential of Data Decision Intelligence in helping companies take advantage of two valuable resources that are more readily available today than ever: access to large volumes of information and powerful processing tools. The board is set, and the pieces are ready to move. What will your next move be?

Florencia Lasa
Director at Tandem.
fl@tandemsd.com