5 Use Cases – How artificial intelligence is already providing support in MDM and PIM projects

We first came into contact with AI in our projects several years ago in the area of automated text creation. Algorithms that were used in journalism, e.g. for generating sports results or economic news, passed our tests for creation of long item texts. In e-commerce we have items with a large number of features. These features enable excellent, target group specific long texts to be generated.In addition to products, dealers receive a large amount of other product information from their suppliers. This has no uniform structure and is not based on uniform classifications or editorial quality criteria.
 
In addition, product information from the relevant manufacturers is supplied in different formats. Therefore, very efficient organisation is needed, along with an appropriate and error-free process for guaranteeing effective data onboarding in a PIM system. This results in a significant amount of manual work, which leads to huge costs. Seven figure sums are far from rare. In simple terms, the ratio of work to benefit is extremely unfavourable.

Our current top use cases for using AI in Master Data Management and Product Information Management:

1. Automated keyword extraction

AI can provide support in avoiding costs and simultaneously increasing data quality. But how might that work? AI can read and extract relevant keywords from the product information supplied and automatically assign them to the PIM classification structures for the relevant products/items. The system takes into account the PIM default values, which are derived from features, flags, attributes etc. Non-unique results are decided by the PIM user – thus the AI learns from each additional data record (“machine learning”). Manual work is substantially cut and there is a sustainable improvement in data quality.

1. Automated text generation

That’s not all. Systematically organised product information, features, attributes etc. enable editorial texts to be created automatically using AI. Based on preset phrasing patterns, AI can use grammatical laws and idioms to independently create product and/or item descriptions based on the information contained and managed in the PIM system.

1. Synonyms etc.

There are limits on the extent to which language is based on multipliable, constant patterns, as (even for editorial texts) it adapts to the relevant situation to convey relevant content (in this case product information). Automatically created texts can quickly come across as monotonous. To address this issue, AI systems can counter the problem during the onboarding process. As part of automated keyword extraction, synonyms, antonyms, hyponyms and other lexical features can also be identified. These are then available for use in automated text generation. This results in a linguistic tone that will appeal to the reader or buyer in the same way as is possible using traditional editorial techniques.

1. Match & merge – Avoiding duplicate item master data

All your efforts were for nothing if you subsequently discover that a painstakingly entered item is already in the system in a modified form – but essentially the same item. This duplicate item means additional item management work and inconsistent item information, as well as falsified reporting as two values are reported for the same item.

With an automated query process during product onboarding, the data manager can be notified of the existing item and duplicate item master data can thus be avoided. As a consequence, you have more time for the items that really require your attention.

1. Market monitoring – the key discipline

Let’s assume for a moment that the facilities discussed above have been implemented and are working. You can assume that the quality of the product information is excellent and that the information leads to sales. But how reliable is this assumption? And how can success be measured? AI systems also provide support here by identifying the relevant issues and trends in the market and taking them into account in the scenario mentioned above. This is done by automated extraction of relevant keywords from social networks, blogs, online magazines, news etc. But AI can do even more than that, as the crucial thing is to correlate the information obtained with actual business data from your own company and then use Business Intelligence (BI) to make it usable.

Summary: Automation is not automatically a rationalisation programme.

The use of AI can automate processes, making them more systematic and reliable. However, this does not necessarily lead to capacity savings. On the contrary in fact. Because of the situation outlined above, available capacity is currently unable to perform its actual roles, e.g. strategic marketing, ongoing development of communication, opening up new markets etc.

The recommendation is therefore that AI can support you in getting back to focusing on your core business.