80 to 90% of a company’s data is unstructured. Hitherto untapped, this wealth of content is about to be unlocked thanks to the potential of generative AI. Gaëlle Helsmoortel, Data Strategy Director, Micropole, explains.

In the field of artificial intelligence and automatic machine learning, the proliferation of data-driven applications is remarkable. However, the challenge lies in managing data that defies conventional structuring, existing in diverse and unorganised formats.

“In companies, an immense amount of information has been collected for years, sometimes even decades… and the vast majority of this data is underused, or even not used at all. This is unstructured data,” explains Gaëlle Helsmoortel, Data Strategy Director at Micropole. You have written documents, customer emails, support tickets, opinions left on the various platforms set up for this purpose, reports from sales and technical sales staff… These are all huge opportunities to analyse ‘the voice of your customers’. In other words, to understand how they talk about your products, your services, your company…”.

But that’s not all: this unstructured data is also an opportunity to anticipate potential future problems, which are still manageable today. On the other hand, if left untreated, it could become a financial sinkhole…

Unstructured data to understand the whys and wherefores

While structured data can provide insight into customer behaviour, or the ‘what’ (names, purchase history and geolocation, for example), unstructured data is better suited to providing businesses with a better understanding of their customers’ intent and behaviour: the ‘why‘ and the ‘how‘.

It’s obviously interesting, but complex. We’re talking about gigantic volumes here. Unstructured data accounts for 80-90% of existing and continuously generated data. The inherent variety of unstructured data also presents an association challenge: how, for example, do you cross-reference images, videos and text? What’s more, the quality of unstructured data is inconsistent, partly because of its variety. Unstructured data can contain errors, inconsistencies or irrelevant information, which can make it difficult to obtain accurate information. Pre-processing or cleaning unstructured data to improve quality can be a time-consuming and complex task. Another concern is analysis. Unlike structured data, which can be queried and analysed quickly, unstructured data often contains a lot of text and does not fit neatly into a database.

Unstructured data is stored in its native format and is only processed when it is viewed. There are also security and confidentiality issues, as unstructured data can contain sensitive information. Finally, integration. Necessary for an overall view, it can be complex due to the absence of a predefined data model.

Customer analysis is a use case for unstructured data

“Today, artificial intelligence helps to extract patterns and therefore meaning from the vast quantities of unstructured data that are created every day”, observes Gaëlle Helsmoortel.

Content that was previously difficult to access within the organisation can now be made instantly useful and valuable, helping decision-makers to discover and drive action, create new content with the voice of their customer and automate processes on a scale never seen before.

Unstructured data is a goldmine of marketing intelligence,” says Gaëlle Helsmoortel. With the ability to rapidly analyse huge quantities of data and personalise customer behaviour, you can better understand what your customers are saying about you, because they are having the customer experience, not necessarily the one you would like them to have, but the one they are really having. And you can really analyse and understand what they want and, as a result, adapt your offers and services. In short, how, in the final analysis, to meet their needs and gain a real competitive advantage!

All the feedback from… real customers

For decades, marketing has wanted to understand and analyse this. It does this by means of studies, which are often very expensive, but they are always probabilities based on a more or less representative sample.

Gaëlle Helsmoortel adds: “Here, we have all the feedback from real customers. You have a huge amount of qualitative, unstructured data that is very difficult, if not impossible, for a human being to analyse quickly. That’s where AI – and generative AI in particular – comes in as the solution! They can analyse a gigantic amount of data in record time and transform it into usable insights, making it easier to make decisions and innovate!

How to manage and analyse unstructured data

By its very nature, unstructured data has no predefined structure that can be easily managed and analysed. So, to analyse unstructured data, you first need to manage it. Start by making an inventory of it, to see where it is. Then define the business objective. Is it to improve the customer experience? Identify product problems? Improve communication through blogs and posts? Or optimising the sales pitch? Finally, analyse the data that will be useful in achieving the objective: easy access? Quality? Quantity? Type of confidentiality?

The choice of GenAI model (open source or proprietary) will follow. Then comes the hard part, with prompt engineering, fine tuning, integration and deployment…

The applications can go very far. Last September, Coca-Cola launched the Y3000, the first can of Coke co-created with generative AI. The first edition is distinguished by its original colours and flavours. The process was carried out in two stages: first, an algorithm analysed data on more than 200,000 existing or potential Coca-Cola products, as well as current and future trends in taste and preferences. Next, the AI proposed several possible formulas, from which the Coca-Cola teams selected the one that best matched their vision of the future.

Generally speaking, generative AI will occupy a key position in improving customer service and boosting user engagement. It will not only improve the customer experience, but also maximise the return on investment in customer relations. Generative AI will make it possible to detect very early on whether a problem exists with one of the products or services of several customers,” comments Gaëlle Helsmoortel. Often not visible because it’s still too early, a detailed, large-scale analysis will make it possible, for example, to detect the repetition of the same remark over a specific period of time and, consequently, to begin a more detailed study of the reasons for it.  

The potential is enormous.“By making the most of this amount of underused data, we will not only be able to save precious resources, but also proactively align products, marketing and digital content with real customer preferences.”