Home PublicationsData Innovators 5 Q’s for Ségolène Martin, CEO of Kantify

5 Q’s for Ségolène Martin, CEO of Kantify

by Eline Chivot
Ségolène Martin

The Center for Data Innovation spoke with Ségolène Martin, chief executive officer and co-founder of Kantify, an artificial intelligence consultancy based in Brussels. Martin discussed how Kantify uses the various solutions to help companies manage customer defection and shared her views on tech entrepreneurship.

Eline Chivot: What gave you the idea to start Kantify, and what is its purpose?

Ségolène Martin: As you know, there is a buzz around artificial intelligence, but AI is a real challenge for executives. It is challenging to know what is AI and how to use it, and to know how to translate a business objective into a performing AI application.

We started Kantify as on the one hand, we had expertise into developing AI solutions, and on the other hand, we saw that companies needed real AI experts to take advantage of the AI revolution. Our mission is to help companies embed AI in their business, and to grow thanks to AI. Kantify is a Belgium-based artificial intelligence consultancy that develops tailored AI solutions for businesses, from scaleups to Fortune 500 companies. We combine management expertise and AI expertise. Our AI performance has already been recognized a few times by major tech players and scientific journals. Our core fields of expertise is AI for sales and marketing and AI for accounting, finance, and procurement. We are mostly active in Belgium, Germany, and France.

Let’s give an example. Let’s say you are a company with B2B or B2C customers, and you lose customers on a regular basis. This is very painful and very expensive for a business to lose customers. Worse than that, it is 25 times more expensive to gain a new customer than to retain an existing customer. Losing customers can happen in many ways: your customers end their contract, or stop buying from you, or buy much less than before. This is called churn. One of our fields of expertise is to help companies preventing churn thanks to AI. By using our AI churn prediction model on our clients’ data, we are able to forecast which customers will leave, for which reasons, and what are the areas where our clients should focus in order to retain their clients. Our AI model can find patterns about customers and churn that no human would be able to detect, so companies can take targeted actions and have a quick return on investment.

Kantify develops tailored software that adapts to the data held by our clients and can seamlessly integrate with their processes or systems. Obviously, we help our clients define what they need and what is the best solution for them. Some of them like to start with a proof of concept, in order to validate their use case and get unique insights from their data, before moving further towards a full deployment. This is pretty common.

Our AI solutions include a number of components. First, the major component is a dataset, or data pipeline (also called ETL or “extract, transform, load”). This is the data that you will feed to your AI model so it can learn. Without clean and structured data, you will get bad results. Setting up a good dataset or data pipeline is a very important element of an AI solution. The second component is the AI model, also known as algorithm. An algorithm is a mathematical formula that makes a prediction based on your data. For example, in the case of churn prevention, the prediction would be: which client will churn, and when. In the case of personalized marketing, it would be: which client will be interested in which product. And in the case of sales prediction, which number of products you will sell in one or six months, for example. The algorithm is trained on the dataset. It will learn from this data and detect patterns. The more data, the better your model will be. This is why we speak about “machine learning.” On top of this, you can add optional components, such as continuous learning. We can train your model so it will learn from the delta between the model’s prediction and what actually happened. Thereby, the model can improve further. This feedback loop is called reinforcement learning. Another optional component is the visualization interface, as for managers and users, it is sometimes necessary to have a dashboard to visualize the data. And finally, an API—an application programmatic interface enabling our solutions to seamlessly integrate with our clients’ existing IT infrastructure.

Beyond helping businesses, Kantify and other AI companies also impact people. AI is seen as a futuristic technology, but AI affects and facilitates people’s lives in several ways already, today, whether we know it or not. When we shop on ecommerce websites and see relevant products, this is thanks to AI. When a fraudulent transaction on our card is detected and stopped so we don’t lose money, this is thanks to AI. When a DNA test enables a doctor to identify if a pathology is due—or not—to a genetic mutation, this is thanks to AI. When our search for images on search engines brings us relevant results, this is thanks to AI.

I am not naïve. As any technology, AI can be used in good or bad ways. But the more I learn about AI, the more I see how much it helps our lives. Each week, we publish a series of “AI reads” with the latest AI news and fields of application. For example, the use of AI for health is just formidable. In cardiology, oncology, genetics, we all know someone who has or will benefit from AI. I could tell more about that as we have developed an algorithm with a cardiologist in our internal AI lab, but this is another story.

Chivot: Tell us about your invoice recognition solution. Which variables and data points do you use?

Martin: Document recognition and especially invoice recognition (also known as invoice parsing) is a field where AI has formidable power. First of all, let me explain what it is about. All companies receive invoices and need to encode the content of these invoices in their accounting system or enterprise resource planning (ERP) for bookkeeping purposes. This is a compulsory task, which is quite repetitive and time consuming. Some companies do the encoding manually, or use optical character recognition (OCR) to extract content from the invoices they receive, such as value-added tax (VAT) and amount.

The problem with traditional OCR-based invoice parsing is that the solutions are very limited. In order to detect a VAT code, they need to know exactly where to look for the VAT code on the document. If the document has a slight change, the solution is lost. Some accountants, fiduciaries, and entrepreneurs don’t trust these technologies and prefer to do this work manually. Kantify has developed a solution, Fyn, that uses AI to understand any invoice, whatever its language or format. With the same accuracy as a human, Fyn can read a document and understand where a VAT, address, discount, etc. are located. Automatically, Fyn can transform data in UBL (Universal Business Language) format. This would not be possible without AI and this is called template-free invoice parsing.

Coming back to your initial question on data, for this, we use a raw input which is an invoice. An invoice is a document which can be paper or digital. It can be in PDF, JPG, HTMLformat, etc. Also, invoices are highly unstructured documents. Even if there is always the similar kind of information, it can be anywhere in the document. Hence why only AI can make sense of it. This is a real technological frontier we are crossing for the accounting industry and robotic process automation in the financial sector.

Chivot: Can you give an example of a project which reflects best how AI and machine learning particularly unlock the value of data for businesses and help them gain a competitive edge in their sector?

Martin: What I just mentioned above is a real good example of how AI unlocks the value of data. One thing is of course that our solution Fyn helps companies to spare money on administrative, repetitive tasks, and dedicate the gained time to some other value generating activities. But the other aspect is that companies will now be able to have much more valuable data. As, thanks to AI, we are able to retrieve and classify the data present on their invoices, this opens the door to new insights and data. Companies can now have better insights on their spend, on the detailed cost of all the items on an invoice (instead of a global cost). This data is gold for financial managers and procurement.

Another example is in personalized marketing (also known in the AI world as recommender engines). Many companies come to Kantify and ask: what can I do with my data? I know much about the preferences of my B2B customers or B2C customers, but how can I turn this into value? The answer is that AI helps turn this data into value in two ways. First, by using the data to personalize the journey of your customer, you can increase your sales and profitability. Then, you improve the experience and journey of your customer, and therefore the value you provide to your customers.

Chivot: Kantify has developed several projects using predictive analytics—to predict diseases for the healthcare sector, price evolution of materials, and customer churn. Can you tell us more about some of the interesting insights such tools generate in these sectors?

Martin: This is a very good question. Many people think “automation” when they think about AI. But in many companies AI is here to provide insights, or recommendations for actions. In other words, AI augments the work of your staff, but does not replace it.

Let’s take again the example of churn prevention. Predicting who will churn cannot be done by a human, in a data-driven way and in a large scale. A churn prevention model will be able to predict who will leave, for which reasons, so sales and marketing managers can take the best decisions in terms of marketing, customer support, and so on. Eighty-four percent of companies think of AI as a means to provide competitive advantage. Having better insights than your competitor is a definite way to get this edge.

Chivot: Kantify isn’t the first startup you founded. What are the opportunities and the challenges for entrepreneurs in the European tech sector today?

Martin: There are plenty of opportunities and challenges! First of all, stepping out of a corporate context has helped me have a new look on management and business problems. As a startup, you need to be agile and adapt to the needs and feedback from the market. This helps you focus on what is important, and find creative solutions to your challenges and the challenges of your clients. But when your client is not able to be as agile as you, due to a very vertical hierarchy for instance, you need to find the best way to communicate and cooperate so we can both learn from each other and do what we both want to do: reach rapid results and value.

Second, being an entrepreneur forces you to learn even better and faster than before. I was already a learner before starting Kantify, but the learning curve has been exponential since I started. You need to learn about many fields, in finance, legal, or marketing, and you can’t afford to be average! You are in the driving chair, and you are responsible for your success.

Third, we are now entering a revolution of the workplace and a revolution of work in general. More and more people are setting up, or considering to set up businesses. It is really fantastic to be part of this revolution and build a business in a technology that is instrumental to the world of tomorrow.

The world of tomorrow is a world where AI will be everywhere, as digital is today. What we currently see is that there are mostly men in the AI industry, which is a consequence of the fact that there are more men studying STEM. I see several challenges in this imbalance. First of all, there is a huge demand for AI and data science talents. Having more women on board would help solve this gap. Second, AI developers have to be able to prevent algorithmic bias, which is when an algorithm reproduces the bias that is in the dataset. There was a recent story of Amazon that perfectly reflects this. This is ethically problematic, and bad for the reputation of a company. Having more women and more diversity in the AI workforce will help identify these biases and prevent them.

I recently launched the Belgian branch of Women in AI in Belgium, as Ambassador of Women in AI—which is a global, bottom up network of AI experts in business, academia, and education. We have ladies from Facebook and the likes, as well as startup founders, university professors, and think-tank researchers. This is a community of people who promote existing women experts and inspire or train others. Women in AI partners with corporates for some of its activities. At the Belgium launch event, I met so many inspired ladies, sometimes leading figures of the AI academic and business world, sometimes with a solid foot in data, sometimes interested to learn more about it. My message to them was: there are plenty of opportunities to be created with AI. If you are not a developer, you can still find or create a job in the AI sphere, or promote AI within your organization!

You may also like

Show Buttons
Hide Buttons