PIM + Artificial Intelligence (AI): the what, the why, and the benefits of a virtuous circle

Artificial Intelligence (AI) has been around longer than you might think. It’s been 21 years since the Spielberg movie ‘A.I. Artificial Intelligence’ helped to popularize the term, but the field of AI was formally founded back in 1956. Progress has dramatically accelerated, with many AI technologies expected to be mainstream within five years, according to Gartner. Their senior principal research analyst, Shubhangi Vashisth, observes “Innovations including edge AI, computer vision, decision intelligence, and machine learning are all poised to have a transformational impact on the market in coming years.”

The terminology in this field can be confusing for less technical audiences. The term AI is also often, and sometimes inaccurately, used interchangeably with Machine Learning (ML) which further confuses the picture. Before we examine what this has to do with digital commerce, and how PIM + AI complement each other, let’s get familiar with all the abbreviations and definitions.

breaking down AI terms

• AI – This is often described as the umbrella term. It relates to the overarching concept of making machines smarter. Aka artificial intelligence.
• Machine Learning (ML) – As a subset of AI, referring to a machine’s ability to learn based on data and algorithms to train a model. Ultimately, the value of ML is its ability to enable machines to self-learn and in doing so create opportunities from data. It’s fair to say most AI leverages some form of ML. Why? The intelligence comes from learned, data-oriented behavior.
• Computer vision – This field of AI covers a machine’s ability to derive meaningful information from digital images, videos, etc. Using that visual input, the machine can act or make recommendations. An example of this is Custom Vision, which is part of the Microsoft Azure Cognitive Services.
• Natural language processing (NLP) – Another branch of AI, NLP gives machines the ability to understand, interpret, and manipulate text and the spoken word, in much the same way as humans. Natural language understanding (NLU) and natural language generation (NLG) are both components of NLP. As with ML, NLP requires structured data for machines to process it efficiently and accurately.
• Generative pre-trained transformer 3 (GPT-3) – This is a language prediction model frequently described as the most powerful and advanced machine-based language tool built to date. What does that mean? It can produce human-like text, for instance, chat convincingly, answer questions or even write poetry.

how does AI affect digital commerce?

With Forbes projecting USD 7.3 billion will be spent by retailers on AI based technologies in 2022, and Meticulous Research predicting AI in retail will reach USD 19.9 billion by 2027, it’s safe to say AI’s making waves in commerce. AI can help brands create personalized experiences on and offline, and it can also help organizations automate and streamline their operations. Here are just a few examples of how AI is fast becoming the backbone of an optimized buying journey.

automated content creation

ML algorithms, when trained correctly, can reduce the manual effort and cost associated with gathering your PIM data. Inriver has long been integrated with tools like IBM Watson that can scan your product descriptions and automatically populate keyword fields from them. Today, we also have integrations with companies like Textual and tools from Google and Microsoft. These may use NLP to analyze product imagery and pre-populate attributes in your PIM, for example, by identifying the color, neckline or sleeve length in a garment, or determining whether a picture shows the left or right shoe.

For longer tail products, where the cost of manual data entry would not be justified by the level of sales, ML can even be used to fully populate the PIM record and get those products onto the digital shelf.  NLP algorithms can also improve consistency, for example flagging a description that refers to “ladies’ shoes” when your company standard is “women’s.”

personalized product recommendations

Recommendation engines are typically what come to mind when talking about AI and digital commerce. You know the one – “other customers also bought” or “you might like…”. Here, brands rely on ML to capture and analyze a customer’s purchase history and that of others, so that it can deliver a personalized experience. In fact, one of our partners, Apptus (a Voyado company) won the 2019 inriver PIMpoint Summit Hackathon with their solution.

conversational commerce

A term coined in 2014, but best defined by Chris Messina in 2016, conversational commerce is “about delivering convenience, personalization, and decision support while people are on the go, with only partial attention to spare.” What does this look like in reality? Think virtual product advisors, messaging apps, live video chat, chatbots, or voice assistance. NLP helps brands to offer a scalable, virtual digital assistant solution around product discovery, recommendations, or even simple customer service questions. Not only does it provide valuable opportunities for upsell and cross-sell, but it helps build an ongoing relationship. In fact, Forbes reported that in 2019, chatbots increased sales by 67%.

digital shelf analytics

AI-powered smart search technology is fast becoming a must-have to thrive in digital commerce. With digital shelf analytics, brands have data-driven insights that enable them to eliminate the guesswork. Engagement intelligence helps brands keep an eye on how their products are displayed, alerts them to stock issues, and allows for visibility to see if their customers are inspired, and converting. It can also provide pricing insights that allow brands to optimize their pricing strategies against their competitors. Inriver Evaluate is an excellent example of such a solution.  

the virtuous circle of AI and digital commerce

So, is it time to automate all our product information? Not yet! We still advise brands to add a human validation step before publishing any machine-derived data. The good news is that the feedback from this human validation helps to train the algorithmic models and make them more accurate. It’s a virtuous circle. The more you use these tools, the more useful they become and the more you’ll want to use them.

It’s also important to remember that algorithms are most efficient when they’re working with highly structured data. Actually, one of the biggest problems for training ML models today is clean and structured data.  For best results, especially from the most cutting-edge tools, you need to store your product data, including images, videos, and other rich media, in a systematic and logical structure. This will help the bots to make connections that are relevant to your specific field of commerce and become exponentially smarter and more useful. Thankfully, creating structure and order within your product information is exactly what a PIM does for you. Adding ML tools to your PIM puts you on the fast track to the virtuous circle and a more futuristic approach to digital commerce.

Thanks to inriver’s AI for drafting my first blog. Not sure if I’m joking? Contact me to find out.