1. Improving Data Quality in Hospital Discharge Reports

Health check systems in Italy are overseen by regional and local health authorities, who actively monitor and regulate the quality of healthcare services to ensure their appropriateness. Over time, numerous Italian regions have developed and revised guidelines and operational procedures aimed at scrutinizing hospital discharge reports and medical records.

The significance of accuracy in medical records cannot be overstated, as errors can lead to various repercussions, ranging from minor billing discrepancies to critical issues such as incomplete or incorrect diagnoses, or delays in scheduling surgical interventions.

In collaboration with Deimos, Rulex leveraged their eXplainable Artificial Intelligence (XAI) technologies to automate the scrutiny of coding in hospital discharge forms within the Alto Adige health authority. The primary focus of the study was to assess the feasibility of applying automatic checks, characterized as logical clinical checks, not only to ensure compatibility between sex-diagnosis or age-diagnosis, as traditionally done with formal logical checks, but also to explore the intricate relationships between clinical variables in hospital discharge reports. This approach aimed to automatically identify inconsistencies among diagnosis, surgery, medical procedures, and Diagnosis-Related Groups (DRGs).

The tested methodologies yielded promising results. Validation rules were defined, resulting in improved efficiency in automatic record checks and identification of probable location of errors, the personnel time required for record checking was significantly reduced, and automatic checks were carried out on all surgical hospital discharge records, not only a test subset.​ Overall, the innovative approach not only enhanced the precision of existing checks but also introduced a more comprehensive and nuanced evaluation of the relationships within medical records.

2. Tailoring Diagnostic Predictions for Primary Biliary Cholangitis

Precision medicine seeks to customize the diagnosis, monitoring, and management of individuals based on their unique genetic and environmental backgrounds. This undertaking is particularly challenging due to the intricate nature of medical traits and the presence of multiple variants. The complexity is further amplified when addressing rare diseases, where limited historical data poses an additional hurdle.

In collaboration with the medical departments of Milano-Bicocca and Humanitas universities, Rulex conducted a pioneering study to assess the feasibility and precision of predicting the risk of Primary Biliary Cholangitis (PBC) using eXplainable AI (XAI). The focus was on identifying novel patient subgroups, disease sub-phenotyping, and risk stratification.

The XAI algorithm was applied to an extensive international dataset of PBC patients, divided into a training set, with 11,819 subjects, and a validation set, with 1,069 subjects, with a meticulous analysis of key clinical features. The primary outcome was a composite of liver-related death or liver transplantation, assessed through a combination of machine learning and standard survival analysis.

The analysis revealed four distinct patient clusters, each characterized by unique phenotypes and long-term prognoses. These findings represented a pivotal milestone in formulating a targeted treatment approach for PBC. Additionally, they laid the foundation for ongoing efforts in identifying and providing timely treatment for the relatives of patients, confirming the potential of XAI in advancing precision medicine for complex diseases.

3. Identifying Correlations with XAI to Improve Metabolic Control in Type 2 Diabetes

One of the primary goals of diabetologists is to establish an effective metabolic control in type 2 diabetes patients, measured through hematic levels of HbA1c, without causing weight gain.​

The Italian diabetology association used Rulex’s proprietary XAI to extract and rank the factors most strictly associated to reducing HbA1c levels. The study involved vast amounts of raw data, including the medical records of 2 million diabetic patients, and the data collected from medical visits over a 10-year period, with over 137 variables per patient.​

Significant correlations were identified, such as the use of specific receptor agonists, while it was established that HbA1c and weight-gain have different determinants. ​These results lead to more efficient patient care for diabetic patients.

4. Extracting Rules to Diagnose Pleural Mesothelioma

Malignant pleural mesothelioma (MPM) is a rare and highly lethal tumor, with its incidence rising rapidly in developed countries due to past asbestos exposure in various environments. Accurate diagnosis of MPM faces challenges, as atypical clinical symptoms often lead to potential misdiagnoses with other malignancies (especially adenocarcinomas) or benign inflammatory or infectious diseases (BD) causing pleurisies. While cytological examination (CE) can identify malignant cells, a notable false negative rate may occur due to the prevalence of non-neoplastic cells. Additionally, a positive CE result alone may not distinguish MPM from other malignancies.

Various tumor markers (TM) have proven to be valuable complementary tools for MPM diagnosis. Recent studies focused on three tumor markers in pleural effusions: soluble mesothelin-related peptide (SMRP), CYFRA 21-1, and CEA. Their concentrations were analyzed in association with the differential diagnosis of MPM, pleural metastasis from other tumors (MTX), and BD. SMRP demonstrated the best performance in distinguishing MPM from both MTX and BD, while high CYFRA 21-1 values were linked to both MPM and MTX. Conversely, elevated CEA concentrations were primarily observed in patients with MTX. Combining information from the three markers and CE could form a classifier to separate MPM from both MTX and BD.

In this context, the Rulex Logic Learning Machine (LLM) was employed for the differential diagnosis of MPM by identifying straightforward and understandable rules based on CE and TM concentrations. Comparative analyses with other supervised methods, including Decision Trees, K-Nearest Neighbors, and Artificial Neural Networks, revealed that LLM consistently outperformed all competing approaches.

5. Extracting a Simplified Gene Expression Signature for Neuroblastoma Prognosis

The outcome of cancer patients is, in part, influenced by the gene expression profile of the tumor. In a prior study, a 62-probe set signature (NB-hypo) was identified for detecting tissue hypoxia in neuroblastoma. This signature effectively stratified neuroblastoma patients into good and poor outcome groups. Establishing a prognostic classifier was crucial for grouping patients into risk categories, aiding in the selection of tailored therapeutic approaches.

To enhance the accuracy of predictors and create robust tools for clinical decision support, novel classification and data discretization approaches were explored. In this study, Rulex was employed on gene expression data, specifically using the Attribute Driven Incremental Discretization technique to transform continuous variables into simplified discrete ones. This pre-processing step facilitated rule extraction through the Logic Learning Machine (LLM). The application of LLM yielded 9 rules, primarily based on the relative expression of 11 probe sets. These rules proved highly effective as predictors, validated independently and confirming the efficacy of the LLM algorithm on microarray data and patient classification.

The LLM demonstrated efficiency comparable to Prediction Analysis of Microarray and Support Vector Machine, surpassing other learning algorithms like C4.5. Rulex conducted feature selection, resulting in a new signature (NB-hypo-II) comprising 11 probe sets, identified as the most relevant in predicting outcomes. This comprehensive approach underscores the potential of utilizing LLM in the development of reliable prognostic classifiers for cancer patients.

6. Extracting Intelligible Rules in Neuroblastoma Prognosis

Neuroblastoma, the most common pediatric solid tumor, poses a significant challenge as approximately fifty percent of high-risk patients do not survive treatment. The urgent need for improved stratification strategies led to the exploration of new, more effective approaches. Hypoxia, characterized by low oxygen tension in poorly vascularized tumor areas, is associated with a poor prognosis. This study aimed to develop a prognostic classifier for neuroblastoma patients by integrating existing knowledge of clinical and molecular risk factors with the NB-hypo signature.

The focus was on creating classifiers that produce explicit rules easily applicable in a clinical setting. The Logic Learning Machine, known for its accuracy, seemed promising for achieving the study’s objectives. The algorithm was employed to classify neuroblastoma patients based on key risk factors: age at diagnosis, INSS stage, MYCN amplification, and NBhypo. The algorithm successfully generated clear classification rules that aligned well with established clinical knowledge.

To enhance stability, an iterative process identified and removed examples causing instability in the rules from the dataset. This refined workflow resulted in a stable classifier highly accurate in predicting outcomes for both good and poor prognosis patients. The classifier’s performance was further validated in an independent dataset. Notably, NB-hypo emerged as a crucial component of the rules, demonstrating a strength comparable to tumor staging. This comprehensive approach showcases the potential of the Logic Learning Machine in developing a robust prognostic classifier for neuroblastoma patients.

7. Validating a New Classification for Multiple Osteochondromas Patients​

Multiple osteochondromas (MO), formerly recognized as hereditary multiple exostoses (HME), is an autosomal dominant disorder marked by the development of benign cartilage-capped bone growths known as osteochondromas or exostoses. Despite various clinical classifications proposed, a consensus remains elusive. This study aimed to validate an “easy-to-use” tool, employing a machine learning approach, to categorize MO patients into three classes based on the number of affected bone segments, the presence of skeletal deformities, and/or functional limitations.

The proposed classification, assessed through the Switching Neural Network underlying the Logic Learning Machine technique, demonstrated a highly satisfactory mean accuracy. A comprehensive analysis of 150 variables across 289 MO patients facilitated the identification of ankle valgism, Madelung deformity, and limitations in hip extra-rotation as distinctive features (“tags”) of the three clinical classes. In summary, the proposed classification offers an effective system for characterizing this rare disease, enabling the definition of homogeneous patient cohorts for in-depth investigations into MO pathogenesis.

8. Predicting Obstructive Sleep Apnea in People with Down Syndrome

Obstructive sleep apnea (OSA) is notably prevalent in individuals with Down Syndrome (DS), with reported rates ranging from 55% to 97%, a stark contrast to the 1–4% prevalence in the neurotypical pediatric population. However, conventional sleep studies are often uncomfortable, expensive, and poorly tolerated by those with DS.

To address this, a dataset encompassing over 460 observations was compiled for 102 Down syndrome patients. Each patient underwent a polysomnogram, and the dataset included diverse information such as clinical visit findings, parent surveys, wristband oximeter data, urine proteomic analysis, lateral cephalogram results, and 3D digital photos.

Utilizing the Logic Learning Machine (LLM), a predictive model was developed to ascertain the occurrence of obstructive sleep apnea in individuals with Down syndrome. This approach aimed to offer an alternative to uncomfortable and costly tests like polysomnograms.

The LLM classification task successfully identified a predictive model represented by a set of simple rules, exhibiting a high predictive value of 81.5% for negative cases. Additionally, the Feature Ranking task allowed for the identification of the most relevant variables, assigning a quantitative score to their importance in the predictive model. This innovative methodology not only facilitates a more comfortable diagnosis for individuals with DS but also provides a streamlined and effective means of identifying obstructive sleep apnea.

9. Benchmarking LLM Performance on Standard Biomedical Datasets

In this study, we employed Rulex’s Logic Learning Machine on three benchmark datasets related to distinct biomedical issues. These datasets were sourced from the UCI archive, a repository of data used for machine learning benchmarking. The datasets are as follows:

1. Diabetes:
   • Objective: Diagnosing diabetes based on the values of 8 variables.
   • Patient Characteristics: All 768 patients considered are females, at least 21 years old, and of Pima Indian heritage.
   • Cases and Controls: Out of the 768 patients, 268 are effective cases of diabetes, while the remaining 500 are controls.
2. Heart disease:
   • Objective: Detecting heart disease using a set of 13 input variables related to patient status.
   • Sample Size: The total sample comprises 270 elements, with 120 cases of effective heart disease and 150 controls.
3. Donor/acceptor DNA:
   • Objective: Recognizing acceptors and donors’ sites in primate gene sequences with a length of 60 (basis).
   • Dataset Composition: The dataset consists of 3186 sequences categorized into three classes: acceptor, donor, and none.

The performance of the Rulex Logic Learning Machine (LLM) was compared to other supervised methods, including Decision Trees (DT), Artificial Neural Networks (ANN), Logistic Regression (LR), and K-Nearest Neighbor (KNN). The conducted tests revealed that the results obtained by LLM surpassed those of ANN, DT (which generates rules), and KNN. Moreover, LLM’s performance was found to be comparable to that of LR.

Predicting customer churn using machine learning has proved effective in identifying potential churners and developing successful retention strategies for the financial sector.

The process of digital transformation has opened up great new possibilities for banks and credit institutions. Online and mobile banking make it possible to reach every customer with an electronic device, no matter where they are in the world.

But as the market broadens, competition increases, any digital bank may potentially be a competitor. And since they are no longer obliged to stick to local banks, customers have a huge range of options when deciding where to open a bank account.

Furthermore, if a bank’s portfolio is uncompetitive in terms of offer and price, or its services are run poorly, it is likely to lose customers. This phenomenon is called churn.

Customer churn: facts and stats

In a nutshell, customer churn refers to customers who stop using a company’s products or services within a certain timeframe. If it happens over a short period, from several days to a month, we call it “hard churn”. It is known as “soft churn” if it happens between a couple of months and a year.

Customer churn has become a real struggle for international banks. According to recent stats, the annual attrition rate for banks in North America is about 11%. This means that banks spend considerable amounts of money and energy signing up new customers just to balance their books. Moreover, attracting and winning over new customers is extremely expensive, something like 6 or 7 times more than retaining existing ones.

In the digital era, banks must therefore focus their energies on developing an effective retention strategy to keep as many customers as possible. In this article, we’ll discuss what causes customers to leave and how to calculate churn prediction using machine learning.

Why customer churn happens

Customers may leave their bank for various reasons. Unfortunately, according to 1st Financial Training Services, 96% of unhappy customers don’t complain; 91% of them don’t explain why they are unhappy and simply leave. Based on our clients’ experience, we have drawn up a list of the main reasons why people decide to switch banks.

• Poor service

Good quality service is the basis for any solid business. But some businesses don’t understand how important it is until they start losing customers. A recent study reported that almost 9 in 10 customers abandon a firm because they experience poor customer service.

• Poor product-market fit

In the age of online banking, customers are constantly looking for better options. This means that if a bank can’t offer a good range of innovative, affordable products, not only will it be unlikely to find new customers, it will certainly lose existing ones.

• Slightly off-key product offers

Even when banks have a competitive product portfolio, they may not know their customers well enough and end up offering slightly off-key products, thereby lowering customer engagement.

• Difficult user experiences

Online banking websites that aren’t user-friendly and are difficult to navigate are a real pain for customers. Not to mention mobile banking apps that crash frequently, interrupting money transfers and online payments.

Predicting customer churn using machine learning

In today’s highly competitive market, successful banks are those which fully embrace digital transformation. Not only do they provide digital services and products, but they also use data-driven technology to enhance their decision-making process.

By exploiting the full potential of customer data, banks can better understand client behaviors and learn churn patterns from past records. This allows them to predict their customers’ future movements and respond accordingly.

Let’s see how.

• Managing customer data with advanced analysis tools

Banks need to smarten up their analytics if they want quick, accurate insights on their customers. But how? Spreadsheets alone aren’t enough to prevent banks from losing valuable pieces of information. They are ineffective when big volumes of data, from different sources and in multiple formats, are involved.

Equipped with an effective data analysis tool like Rulex Platform, banks can easily merge all their customer data from a wide range of databases into a single place in the same format. This facilitates data analysis, allowing banks to learn client behaviors and implement an effective retention strategy.

• Tracking churn patterns from historical records

Many different drivers cause customers to leave their banks, and they vary across life stages and demographics. Drawing on a bank’s historical data, machine learning models can generate insight on churn patterns. This helps business experts make predictions on possible future churners.

Rulex’s eXplainable AI (XAI) has proven particularly effective in the financial services sector. Rulex’s XAI quickly analyzes historical records and produces explainable outcomes regarding which customers are more likely to churn and why. On high-confidence predictions, Rulex technology’s success rate is between 90 and 99%.

• Increasing customer satisfaction with AI

Banks need to engage with their customers on a deeper level to improve customer retention. Knowing your clients means, for example, being able to offer them on-key products at the right time or enhance their experience. Once again, AI technologies like Rulex’s XAI can come in handy by suggesting the best ways to improve customer satisfaction.

After pinpointing customers who are likely to churn, Rulex’s XAI proposes corrective actions to prevent them from leaving. For example, for customers dissatisfied with service, Rulex’s XAI will suggest improvements to its quality. The corrective action will be to place customers who contact the support center twice or more into a special priority queue, so their calls are handled more quickly.

By using historical records and applying similar types of corrective action for all their customers, one of our clients was able to reduce churn rate by almost 3%, equal to a revenue increase of 11%.

If you feel it’s time for your bank to tackle the issue of customer churn, visit our Financial services page and get in touch with our experts for a free consultation.

How many algorithms do we encounter on a daily basis? The answer is much more than we think – from consulting Google maps to find the fastest route to that new restaurant, to scrolling through our Facebook feed.

But how does artificial intelligence decision-making work? And are all algorithms the same?

Relying blindly on a new technology is extremely risky, particularly when dealing with sensitive data. No one would want an important decision like a credit evaluation or a medical diagnosis to be decided by an algorithm that we can’t understand. This is where the GDPR and other international regulations come into play, requiring that algorithms involving personal information process data transparently and provide a clear explanation of any predictions made.

In this article we will discuss different types of artificial intelligence techniques, from black box to explainable AI (XAI), shedding some light on the subject.

Explainable what?

First, we should ask ourselves what “explainable” means in the context of AI. But before we ponder exactly what eXplainable AI (XAI) is, we should ask what we mean by transparent solutions and considers their explanations. We distinguish between two types of explanation:

To be 100% explainable, AI solutions should provide both types of explanation. The first fosters trust in artificial intelligence, whereas the second explains the reasons behind each decision. The latter is required by law when the automated decisions affect people’s lives.

Shedding light on AI

In general, there is a tendency to divide algorithms into two categories: black box and explainable algorithms. The difference lies in their ability to provide outcome-based explanations. When using black box algorithms, it is impossible to understand the logic that led to the output, whereas with explainable algorithms it is possible to explain both the process and the specific output. But the reality is a bit more complex than that. Some black box algorithms are more explainable than others, and some explainable algorithms are less explainable than others.

Explainable AI

XAI solutions produce explainable predictions, but some XAI solutions are less understandable than others, meaning that only AI specialists can explain the output, after complex analysis. This category includes, for example, regression, linear, logistic, and LASSO algorithms.

Conversely, some types of XAI solution have a very high level of explainability and transparency. In these solutions, the output is expressed as rules (e.g., IF-THEN rules: IF rain & IF work THEN take the car), which are easy for business experts to understand. Among the algorithms that fall into this category, we have created the Logic Learning Machine (LLM), which produces IF-THEN rules and has high performance levels.

The LLM can be used in a variety of business scenarios, especially when a high level of transparency is required by law to protect people’s rights. This happens, for example, when dealing with sensitive decisions like granting loans or detecting cases of fraud. The LLM can also be used to empower a business with high quality data, and to detect and correct data entry errors (read more).

Black box AI

The term “black box” stems from the fact that the predictions are too complicated for any human to comprehend. The output of black box AI may be given by layer upon layer of interconnected computations involving multiple parameters – millions, or even billions. This makes it impossible to trace how the final result relates to the original input features, such as age or gender. The category includes neural networks, support vector machine, and deep learning algorithms.

The problem of black box AI can be mitigated by approximating the model with a more understandable one – “opening the black box”. However, the explanations obtained by the explainable model may be inaccurate in the best-case scenario, and misleading in the worst case, causing issues when they are applied to sensitive use-cases. Among these techniques are LIME, Shapley, and PDP (more info here). They therefore differ significantly from the aforementioned explainable techniques.

Proprietary algorithms are a case apart. They are not black box per se, but the companies who own them hide details of their AI system to protect their business. These are the types of AI we interact with most frequently: Google Search’s ranking algorithm, Amazon’s recommendation system, Facebook’s Newsfeed, and more.

Why businesses choose eXplainable AI

Mindful of the general call for data ethics, more and more businesses are now choosing eXplainable AI solutions. As shown by recent stats, the global XAI market is expected to grow by 513% by 2030, reaching a value of 21.7 billion U.S. dollars. Choosing eXplainable AI therefore brings major advantages for companies such as:

1. Guaranteeing better and fairer decisions
2. Building trust and credibility with customers
3. Complying with the GDPR and international regulations
4. Staying human-centric

To find out more about Rulex’s eXplainable AI and how it can help your organization, take a look at Rulex Platform.

Explainable AI it produces transparent, easily understandable models. Using a series of if-then statements, Rulex automatically produces self-explanatory logic for all decisions. Rulex rulesets make it possible to explain a decision directly to the customer or provide customer service agents with the ability to look up the reason for a decision.

Why eXplainable AI is more transparent than black box?

The problem with conventional AI is very simple: it’s unexplainable. Conventional AI relies on machine learning algorithms such as neural networks and others that have one key feature in common: they produce “black box” predictive models, meaning they’re mathematical functions that cannot be understood by people, even mathematicians.

f(x) = 0.293 tanh(0.337 x1 - 0.329 x2 + 0.251 x3 - 0.288 x4 - 0.297 x5 +
0.436 x6 + + 0.166 x7 - 0.184 x8 + 0.219 X9 ± 0.483 x10 - 0.222 x11 + 0.173 X12 
± 0.012 X13 ± + 0.352 x14 + 0.259 X15 ± 0.176 x16 + 0.345 x17 + 0.314 x18 + 0.177 
x19 - 0.329 X20 - 0.3) + - 1.934 tanh(-0.233 x0 + 0.174 x1 - 0.252 x2 - 
0.501 x3 - 0.125 x4 + 0.311 X4 - 0.573 x6 + - 0.299 x7 + 1.123 x8 + 0.318 x19 - 
1.169 x10 + 0.105 x11 - 0.429 X12 - 0.075 X13 ± 0.143 X14 + 0.146 x15 - 0.531 x16 
+ 0.077 X17 -0.133x18 0.122 xl9)

Rulex’s unique, proprietary machine learning algorithms work differently. Rulex creates predictive models in the form of first-order conditional logic rules that can be immediately understood and used by everybody. Here an example of Rulex clear box predictive model.

IF customer_province in {A, B, C, D} AND damage_class in {1} 
AND Number of days between policy start and date of accident <= 371
THEN Fraud = Yes
IF customer_province in {E, B, C, F} 
AND Customer age > 48 
AND Number of days between date of accident and complaint > 1)
THEN Fraud = Yes
IF customer_province in {G, H, I, J, K, L, M, N, B, O, P, Q, R, S})
AND Number of days between policy start and date of accident > 371
THEN Fraud = No
IF (Number of days between date of accident and policy end <= 2)
THEN Fraud = No

How eXplainable AI works?

Rulex’s core machine learning algorithm, the Logic Learning Machine (LLM), works in an entirely different way from conventional AI. Rather than producing a math function, it produces conditional logic rules that predict the best decision choice, in plain language that is immediately clear to process professionals. Rulex rules make every prediction fully self-explanatory.

And unlike decision trees and other algorithms that produce rules, Rulex rules are stateless and overlapping, meaning one rule can cover many cases, and many rules can cover a single case. This allows for fewer, simpler rules and provides broader coverage at the same time.

Rulex calculates the coverage and accuracy of each rule, making it easy to select the most effective decision rules. Also, proven heuristic human rules can be added to the predictive model, allowing a seamless blend of human and artificial intelligence. Human rules are also rated for coverage and accuracy, allowing Rulex to easily evaluate the quality of the decision rules in use and reduce false positives.

While Rulex is one of the most innovative tools for making true explainable AI, it’s not the only one. If you are curious to learn more, you can read 8 explainable AI frameworks for transparency in AI.

Where is eXplainable AI important?

Explainable AI is important to any business because it conveys trust and competence. It is particularly relevant in some sectors and applications where decisions can have a strong impact on people, such as granting a loan (read more) or making a medical prognosis (read more). According to many international privacy regulations, such as the GDPR, artificial intelligence cannot replace human decision but only support it.

Privacy is one of the areas in which eXplainable AI plays a more important role. Here are some tips for creating and maintaining processes with artificial intelligence that respect the principles of privacy.

5 Tips for Privacy Compliance

1. Identify processes in your business that use profiling and automated decisions.
2. Inventory the machine learning models currently used.
3. Assess your existing models. Are they interpretable? Can you demonstrate to an auditor that they do not discriminate?
4. Assess your current machine learning techniques. Do they produce interpretable rules?
5. Develop a strategy for meeting compliance requirements in each stage of the machine learning workflow

How focusing on decisions can help you productionize your next AI project

The internet is the best place in the world to turn your suspicions into nightmares. Suspecting your partner is cheating on you? Every Facebook Like will confirm it. Feeling a little dizzy? Dr. Google will immediately diagnose how many days you have left to live. Wondering if the Earth might really be flat? Well, you get the idea… Shifting to your workplace:

If you are wondering whether your company should adopt AI, the web will serve you crucial “insight” on the imminence of your bankruptcy if you don’t act immediately.

What differentiates AI from ML?

To clearly understand the difference between AI and ML, I personally like John McCarthy’s definition of AI, as it is very simple:

“AI involves machines that can perform tasks that are characteristic of human intelligence.”

Such tasks include things like understanding natural language, identifying objects in images, recognizing sounds, and playing complex strategy games. I find this definition very powerful as it does not put any stress on the underlying technology. It basically tells us that AI is a glorified version of good-old process automation, which now includes human-centric processes that weren’t possible just a decade ago.

ML, at its core, is nothing but one of the many technologies used to achieve AI. Disruptive, innovative, sexy… but still just a technology. If we don’t untangle this difference, we will find ourselves asking the mother of all incorrect questions: “What problems can I solve with ML?”.

This question unconsciously traps you into searching for the right problems for the technology at your disposal, which is never a good business approach. As the famous psychologist Abraham Maslow once stated: “If all you have is a hammer, everything looks like a nail”.

The problem is that your company might not need a nail at all.

Don’t get me wrong, I’m not saying you should never ask yourself this question. I’m simply saying that this should not be part of any AI-strategy conversation. It’s a headache your data science team will be more than happy to take on.

How to NOT fail

After years of experience in complex supply chain automation (working mostly in these scenarios), I’ve seen projects fail for reasons ranging from using the wrong technology to manipulating dirty data or lack of team cooperation. While addressing these problems is clearly important, not understanding the logic behind decisions and overlooking their impact on the business is by far the most lethal mistake.

Focus on the decisions you want to automate, not on the technology.

Decisions are the natural outcome of any learning process; we learn things to better react to the situations we face and to avoid previous mistakes. At the end of the day, introducing AI in your company is nothing but allowing machines to transform your data into decisions. That’s why, in every successful project, we have always started with “the end” in mind, focusing on the output we wanted to create and asking ourselves: what decisions are we trying to automate? by how much and when do we want to improve the decision-making process we are looking at?

It’s all about Decision Automation

It was interesting to notice how focusing on understanding decision logic led us to become increasingly detached from technological conversations. The term AI basically disappeared and was replaced with “Decision Automation”, which, while not a new concept, isolates the final outcome and its scope of work: enhancing the quality of our decisions and removing humans from the part of the underlying process which does not require judgment, creativity or control.

Let AI do (only) what it can do better than you.

Focusing on decisions can greatly help us build a simple framework that can better identify and tackle our next AI-project.

Start by asking the right questions

Some of the questions we might want to ask ourselves are:

• Are we looking at operational or strategic decisions? Operational decisions happen daily and repeatedly; they are often boring and unedifying for the people in charge. They rely on well-defined rules or logic and are therefore the perfect candidates for automation. For example, saving time while reducing inefficiencies should be the focus of our attention when replenishing distribution centers, identifying fraudulent claims, or non-performing loans. On the other hand, strategic decisions such as “should I make this investment?” or “should I partner with this company?” require unstructured insight, which quite simply cannot be automated, but only, as defined by Gartner, “augmented” by using the right technology.
• What is the impact of wrong decisions? Being able to shape the effect of wrong decision-making, both in terms of lost money and people affected, is essential when prioritizing the tasks you want to automate. Experiencing recurrent out-of-stocks or overstocks could lead you to optimize your replenishment process, while a large part of the problem might be due to an incorrect setup of your master data, which is affecting not only distribution but also production, transportation, and forecasting (true story, by the way).
• Can we decide fast enough to modify events in due course? While the quality of our decisions might be good enough, the process involved to reach these results may be excessively taxing on the business. For example, most business-critical activities in the supply chain are still done manually or semi-manually, and they are consequently lacking in flexibility and resilience.
• Do we know the logic behind the decisions? Do we know how and why something happens? Finally, we talk about technologies! If the answer is yes, technology can provide businesses with support through RPAs (learn more on RPA) and rule automation for simpler tasks, and low-to-no code ETL tools and optimizers for the more challenging ones. If the answer is no, but there is an underlying logic, then ML can dig it out from the data. An example is customer churn analysis, as it is impossible to predict upfront what drives customers to leave, but that information is probably hidden inside the data.

In conclusion

Being able to provide quantitative answers, such as the number of decisions involved, the inherent cost of wrong decisions, or the man/hours needed to support the process, should be the gateway to automation investments.
These answers help us build compelling cases to convince management of its value and serve us as leading indicators, whether we are ready or not to invest in innovation and automation.