Intro to ML

server interior ai

Machine Learning in Business; Why and When?

By Featured, Intro to MLNo Comments

Machine Learning (ML) is a game changer for businesses of all kinds. It is critical for executives to understand more about how it could transform them. To that end, this article focuses on the two key points of why and when Machine Learning should be used.


There are three reasons why a company should use ML:

First, to create economic value: new ML based products can be developed or existing ones greatly enhanced. New intelligent or self-tracking cameras and scanners, recommendation systems for websites or apps, are just a few examples of these ML applications creating value. In numerous cases ML enables the findings of hidden patterns, that even the most trained human eye could not notice. One example can be found in the computational biology field, where ML algorithms frequently outperform doctors. For cancer treatment, there are ML models which recommend which drugs a patient should take based on numerous health features such as their blood types, tests results, scans and more that no doctor could consider as a whole. Another one is recent work done by academia and the major tech companies in the field of computer vision, where state of the art techniques are better at reading road signs than humans.

Second, ML can be used to reduce costs by optimising processes. This is referred to most frequently as automation. Many business processes are redundant or tedious and could be streamlined through using computers: filling in forms, handling client’s requests, filtering information, improving the knowledge transfer, automatically check the quality of products, and more.

McKinsey&Co recently reported: “In about 60% of occupations, at least 30% of activities are automatable”. Below, we show a graphic illustration from the US bureau of Labor Statistics, showing today’s automation potential thanks to AI.

Moreover, in many industries considerable waste is produced daily which could be reduced thanks to ML models predicting more accurately the demand and adapting the production and inventory levels accordingly.

Third, ML can increase revenues by enabling a more throughout market understanding. We see these applications as being mainly applied in marketing and finance. Customer segmentation, pricing or portfolio optimisation are some examples of such applications.

When taking companies from their inputs side (so as entities which buy assets), being able to better understand the value of the goods acquired allows to gain bargaining power or reduce risk. The bargaining power can come from the anticipation of the:

  • supplier’s prices
  • competitions willingness to pay
  • market demand.

ML models can predict this kind of variations efficiently when provided with the right data. For example, in the energy sector or in heavily energy dependant industries, being able to optimize the production scheduling with price forecasts can allow considerable cost reductions.

Now, when taking companies from their outputs side (so as entities which sell assets), being able to better understand the willingness to pay of their customers will allow companies to capture value by adjusting their prices, packaging and quantities produced. ML allows to efficiently target customers and understand which product features bring them or make them think they acquire more value when buying the product or service. 

These were three reasons and some examples that were based on a microeconomics perspective. However, another point should be mentioned before moving on to the “when” part of this article.

ML hits all industries in a disruptive fashion. Due to ML applications’ great return on investment and their high level of scalability, they make the competitive advantage they bring crucial especially for early adopters who will be able to capture an important share of the newly available value.

When should business apply ML?

As we just learned the reasons to use ML, it follows to need to understand when it is valuable to use it. Specific requirements need to be met in order to undertake the development of ML projects. First and most importantly, having the idea of the right ML applications to invest in is a necessity. Second, the right data needs to be available, whether it comes from internal company data or from public sources. And third, making sure the business impact of the project will be positive for the company is needed as for any investment.

Ideation Process

Ideation of ML projects can be difficult when lacking the expertise. It requires considerable knowledge of the existing techniques and the kinds of data that can be used effectively to build various models. On top of that, business expertise is equally critical to come up with high value ideas solving pain points or creating new value for a business.

Several options are offered to companies:

  • Read ML books and papers.
  • Read articles that describe case studies such as the ones we publish frequently ( in order to make connections and see what could apply to them.
  • Contact us, and have us identify their highest-value opportunities in ML.


ML could not exist without data. This data needs to be available in sufficient quantity and quality.

For quantity, the data on which the artificial neural network will be trained needs to be representative of the whole problem for the model to perform well. An ML based quality control image classifier cannot be expected to predict if a piece is defective or not just by showing it one way a piece could be broken. The model needs to have learnt a lot of the possible ways a piece could be defective. The more relevant the data the model will be trained on, the more accurate the model will be able to predict.

Regarding quality, the data labels needs to be correct. This same model will not learn differentiating faulty from working pieces if it is told that working pieces are faulty and vise versa. The data also needs to be as precise as possible. If the images of the pieces to classify are blurry, it will be hard for the model to understand which features make a broken piece broken.

The data can come either from businesses: internal knowledge base, CRM data or client feedbacks, ERP or production data (sensors, inventories, scanners, monitoring systems, etc).

But the data can also come from external open source datasets which are extremely helpful to increase the performances of ML models. There exists many of them but one of the most famous ones is a handwritten digits dataset which allows to train very performant models that are able to read more accurately than humans.

Oftentimes the data can be sensitive. This issue can be handled in multiple ways. The data can be anonymised or encrypted and private servers can be used. This is the case for Visium, which stores its clients’ data on own private servers in Switzerland.


ML’s value needs to be estimated in order to take the strategic decision to use or not to use it. When integrated in businesses, the value Machine Learning applications bring is actually most of the time easily measurable.

ML models are trained and then tested. The testing phase gives an accuracy (or other indications) which says that given new data, it will predict or classify it properly with this measured accuracy.

In the case of economic value creating and increase in customer/provider understanding ML applications, one way to estimate their value is to compute the extent to which the price and quantities are expected to be more favorable and what it will bring versus the cost of the investment.

In the case of cost reduction ML applications it is often simpler as it is enough to compare the cost structure with or without the application (without the application being most of the time the current cost structure). For example, having this model will allow the production line to have its pieces checked with a 95% accuracy at a cost of running a computer and maybe 0.1 second of processing per check. This can be compared to the current cost structure and a precise data driven conclusion can be made.


It should now be clearer why ML can bring value to your business and when ML could and should be used. We hope to have helped you understand that becoming an early adopter will greatly benefit your company and will spare you the risk of being left behind by the competition. Feel free to contact us for a free preliminary consultation about your business needs and potential application of Machine Learning.

If you want to learn more about ML applications, follow our page on linkedin and twitter.

How can AI help your business adapt to uncertainty and drastic external changes?

By Featured, Intro to ML, VideoNo Comments

Flashback to an interview with our COO Matteo Togninalli by INSIDE TOMORROW Innovation & Technology TV talk show and Yelena Ganshof, on how AI and ML technologies can help companies weather uncertainty.

Matteo touches on one of our exciting projects based on Audio Signal Processing. This AI algorithm detects accidents or criminal events in cities and dispatches authorities to the site more efficiently. Find out more about exciting projects like these and the extraordinary ways in which AI and ML solutions can help companies transitions through times of change, watch the full video here

blog corona and ai

5 Ways Data Science can help you during this Crisis

By Featured, Intro to ML, ProductNo Comments
five ways ai can help in crisis diagram

(1) Situation Dashboard 

What is it? NLP enhanced content filtering of key sources as well as customised data aggregation enables  distilled valuable insights about multiple facets of a topic. This data can then be visualised in an interactive dashboard to provide data-driven decision making for leaders. 

Crisis use? To navigate the Covid-19 pandemic effectively, leaders need quick access to the latest information across public (e.g. latest government policies) as well as company internal indicators (e.g. cases among workforce, supplier defaults). With an AI enhanced information tracker, customised information can be gathered, aggregate and visualised effectively to provide decision support. (Could also be augmented with modelling)

(2) Enhanced Forecasting (Financial, Sales, Delays, Workforce)

What is it? Time Series modelling takes into account historical data and exploits mathematical models to predict trends and fluctuations in sales and financials, supply chain delays as well as workforce needs. At Visium, we have enhanced forecasting with risk management frameworks and the possibility to analyse multiple scenarios generated with probabilistic programming. This approach enables the identification of insights and accurate forecasts in times of uncertainty and rare events.

Crisis use? While the effects of COVID-19 are playing out, predictive tools can help navigate the new normal and guide a timely response by giving you information on what to expect in your supply chain, sales and workforce. With our approach, we can model of possible alternative scenarios and empower business decision making in response to new, unprecedented black swan events.

(3) Automated customer care 

What is it? This chatbot-like solution can match customers questions with a set of continuously updated answers. Our solution differentiates itself from normal chatbots through its ability to intelligently query a large number of documents and extract answers, even for novel questions. This solution allows for customers to more easily find the answers they are looking for which in turn increases customer satisfaction. 

Crisis use? During this time of increased uncertainty, customers reach out to you more frequently, which can overextend your employees capacity. Visium’s automated customer care can help customers find a satisfactory answer online rather than calling or e-mailing. 

(4) AI enhanced contract management 

What is it? This NLP tool extracts and classifies information within contracts, including both digitally produced as well as scanned documents. This allows you to then filter out the information you need, or classify the documents based on the relevant information. 

Crisis use? In the current situation, AI enhanced contract management can automatically investigate thousands of contracts for the implications of the COVID-19 outbreak on contract clauses of any kind (e.g. compensations in favour of any party or clauses affected by pandemics).

(5) Supply Chain fortification 

What is it? Using state of the art Graph Machine Learning and Monte Carlo modelling techniques, supply chain bottlenecks and vulnerabilities can be identified.

Crisis use? As companies navigate the Covid-19 crisis, the importance of a reliable supply chain becomes ever more apparent. Visium’s unique capabilities can help guide the fortification of supply chains and identify opportunities to increase efficiency.


    blog what are AI - ML

    What are AI & Machine Learning?

    By Featured, Intro to MLNo Comments

    Humans are good at a lot of tasks. For some of the tasks we learn, we have a clear procedure in mind. For example we know how to multiply two numbers as we have learned in school as children the exact steps we need to take to accomplish the task. We can use these exact steps to instruct a computer how to multiply two numbers. With these instructions the computer can do this particular task much faster than humans, billions of numbers can be multiplied in a second on today’s computers.

    On the other hand there are tasks that we, as humans, can do easily without having to think of a clear procedure.  An example of such a task is the infamous problem of distinguishing cats and dogs based on pictures. The interesting difference of this task compared to multiplying numbers is that we do not have an exact procedure on how to do it. How do you distinguish a cat from a dog? Do you look at the size of the animal? Do you look at the color of the fur? Do you measure the length of their fur or tail, or look at the shape of their ears? How do you manage to still tell dogs from cats if you only see just their head, or even only one of their limbs? Likely you will answer, I don’t know precisely – but a cat just looks different from a dog… Throughout our lives we have seen many cats and dogs and built a mental model of what these animals are supposed to look like. The problem is that if we want a computer to do this task, we have no clear instructions on how to solve the task.

    How do you teach a computer a task that you do not know the steps to succeed? We try to imitate nature. Just like we were inspired by birds to build aircrafts, we would like the computer to learn the task based on examples similar to how we do it ourselves. Instead of a mental model, we specify the structure of the model as a black box function with millions of variable parameters. The system then learns from examples and self-adapts these parameters to better and better solve the task at hand. This self-adaptive system proves to be successful again and again for tasks that were previously thought to be unsolvable.

    how ai works cat example

    How do computers learn?

    The simplest approach to incorporate knowledge into a computer algorithm is to let experts define explicit rules. For example we can look at the height of the animal and if it is below 25cm, we decide for a cat. We can also combine multiple criteria for more accurate results. Such rule based systems are considered Artificial Intelligence.

    How do we come up with these thresholds? It would be much easier if we can learn suitable thresholds from the data. Automatically learning rules based on a set of features brings us to Machine Learning. These features can be hand-designed, e.g. the color of the fur or the length of the tail. Typically domain experts define features that make sense. However, extracting such features can be tricky depending on the domain and often it is even unclear what suitable features are. It would be much more elegant if there was a way to even learn the features directly from the data.

    Today we have the power to achieve automatic feature extraction with Deep Learning. Using large collections of images it would allow us to learn to distinguish cats and dogs with high accuracy. Unfortunately this approach has its drawbacks. The availability of large amounts of data is not feasible for all domains. Furthermore the outputs of these models are often not explainable. We started using this approach since we have no way of describing the exact steps that can solve the problem, so expecting we get a clear explanation is a bit paradoxical. There are ways to sacrifice performance to get better interpretability for domains where it is a necessity.

    deep learning machine learning ai blog

    What is Machine Learning?

    So what is machine learning apart from a trendy buzzword? In a nutshell, machine learning is a set of statistical tools that enable data-driven informed decision making. These tools allow us to transform data into knowledge.

    Why now?

    In the past few years the field has gained a lot of interest, both in academia and the industry. There are multiple reasons for its emergence: 

    • Data availability
    • Processing power

    The first reason is what was called The Age of Big Data a few years back. Many industries have moved to digital processes, sensors keep track of the world around us and we decided to store all of it. The difference is that now we not only have the capacities to record and store massive amounts of data, but we also have the processing power to actually extract meaningful insights from the data. With the advances in processing power of the past few years it is possible to run these models at scale in production.

    All of this has led to the revolution we see today. For many tasks these machine learning models outperform humans. It can be in the game of chess, image classification, cancer detection, inventory forecasting and many other domains.

    Cookies / Privacy Policy

    Visium 2020
    Developed in Switzerland