Without a doubt, Artificial Intelligence is the number one story of 2023, with breaking news and hyperbolic predictions popping up daily. Retailers and their partners are justifiably concerned. Retail marketing and advertising directors are asking what AI will mean for their current operations and the future of retail marketing in general.

The first thing to remember is that AI and its related technologies – machine learning (ML) and “big data” – are nothing new. The idea of using large training data sets to automate routine activities or predict future outcomes, often called “narrow AI,” has been around for decades. But with all the advances in cloud computing and the November release of OpenAI’s ChatGPT, the floods of new information (and misinformation) have created more than their fair share of anxiety for some retail professionals. However, the secret to overcoming these fears is the same as it has been for any new technology – understanding its abilities and limitations.

Basic concepts

AI can be applied to almost any digital data, including structured data (usually alphanumeric fields with logical labels and relationships) and unstructured data. Think of structured data like a spreadsheet with clearly labeled rows (records) and columns (fields). Unstructured data does not have such neat labels and includes digital images and unclassified text, such as that found in social media posts. But even if a data set doesn’t have logical, structured labels (or metadata), it can still have meaning.

AI is designed to detect patterns in large volumes of unstructured data. Those patterns, often confirmed with human assistance, are used in ML algorithms to train a system to recognize similar patterns – in images or text – and make predictive decisions automatically. An AI does not understand the meaning of those patterns as a human can, but it can simulate a human’s meaning-aware response far faster than human decision-makers.

Let’s consider the difference between human understanding and how AI can simulate it. A retail marketing manager, for example, may understand the meaning of certain facts about a product:

1. It has sold well at certain times of the year.
2. It has a reasonably high-profit margin and a reliable supply chain.
3. The manufacturer has supplied most of the product’s relevant information.
4. It has had many favorable reviews on social media and elsewhere.
5. It has been favorably described in various blogs and articles.
6. There are publicly available images and videos of people using it.

Items 1–3  represent structured data, typically found in sales history, inventory, or product information management (PIM) databases. Items 4–6 are unstructured, for the most part. But from those data points, a human might reasonably conclude that a sales campaign for that product is a good idea. But because there are so many different products and product variables, it would be impossible for one human being – or even an entire marketing department – to make those decisions on a massive scale. On the other hand, an AI-based system can detect meaningful patterns (as confirmed by humans) from all kinds of available data and automatically prioritize the most likely candidates for a marketing campaign.

Doing this at scale would accomplish several things. If the AI’s pattern recognition is accurate – an easily testable hypothesis – then the effectiveness of retail marketing campaigns would be greatly increased. It would also lessen the cost and drudgery of combing through data to find meaningful, actionable insights. Finally, if marketing and advertising directors were freed from these burdens, then they could focus more on things that artificial intelligence will likely never do, namely consider the aesthetic and psychological preferences and biases of their audience. Creative and insightful humans will always have the edge regardless of how well or quickly AI can mimic human behavior.

The retailer’s advantage

In this area, large retailers already have a big advantage over other companies, namely their ready access to massive amounts of mostly structured, product-related data. This is only natural since they must handle thousands (or millions) of individual products from multiple manufacturers. Every product SKU must have an array of feature and component data, usually stored in a PIM system. At the same time, it must maintain all the images and descriptions of each product, typically stored in a digital asset management (DAM) system. Add an array of other data sources for sales, inventory, pricing, and other essentials. Many companies now also include e-commerce sales histories and customer reviews.

That ocean of data has to be well managed, of course, which can be challenging. Fortunately, Comosoft’s data integration teams have helped many retailers unify their PIM, DAM, and other structured data sources, using the LAGO system for meaningful, collaborative planning and efficient, InDesign-based workflow optimisation for multi-version print and digital campaigns.

Of course, the difference is that the typical PIM, DAM, and other data sources are highly structured – as they should be. We are only in the early stages of adding structure to new varieties of data, such as customer reviews and customer-supplied images and videos. But those new data types offer incredible potential value to the retail marketer. And AI will only accelerate that value.

Focus on data-readiness

Before anyone can realistically tackle AI and masses of unstructured data, they must first master their structured data – turning it into marketing gold, as it were. Comosoft LAGO is a proven tool for doing just that.

Once a large retailer has found a way to be “data ready” with what they already have, the next step into artificial intelligence will be a logical, powerful next step. With this readiness level, AI can only launch the retailer into new levels of efficiency and growth.

Find out more about Comosoft’s custom systems integration and data strategy service and book a demo to see for yourself how LAGO can streamline your data workflow.