How AI is changing the retail landscape (2/2)

The applications and technology behind them

Deepak Singh
4 min readAug 8, 2022


I n the first part of this series, we learned how Computer Vision (Convolutional Neural Network, CNN) is one of the most advanced and useful tools in the AI-for-Retail toolset. In this part, we will see how NLP, Clustering, and Logistic regression (3 remaining key AI technologies), can help accelerate retail development in near future. We will focus more on NLP due to its huge benefits and less exploration done as of now.

Natural Language Processing (NLP)

Let’s start with what is NLP and how it works. NLP stands for Natural Language Processing. Any NLP algorithm first countifies the data (i.e. set of words formed in sentences) and arranges them in a logical format and then grammatical rules are applied to make sense of it. Now, when these rules when turn out to be correct, it provides negative and positive feedback to the system to motivate the system to refine or strengthen its confidence in its rule/formula. NLP can process any kind of data from social media comments to customer reviews. These are heavily used in sentiment analysis in comments, reviews, and chat history. It is first trained with a huge set of data with their actual output to train the model to identify the meaning. Here are some existing applications that we may find around us:

Instore robots for assistance — Furhat Robotics

We will soon see how the robots are assisting the customers, in making purchases, making payments, and sometimes also walking behind the customer while carrying the basket to collect the groceries from different sections. This is one of the very interesting and promising use cases of keeping out bias using a robot.

Chatbots for online customer service — Ada

Chatbots are one of the very early applications of AI in retail. Initially, it used to be very rules-based but now it is getting more intelligent. Ada is one of the top companies that provide its customer experience services through conversational AI. It also provides autocomplete and autocorrect services through a chatbot or separately.

Optical Character Recognition (OCR) — Rossum

OCR is actually a blend of computer vision and NLP. The CV extracts the letters out of the images and then NLP makes sense out of it in multiple ways. In retail, it can help to extract text out of bills, invoices, contracts, packages, and letters from customers and then make sense out of it and send a summary to the retail store managers for further analysis. It actually subtracts the repetitive and mundane work from the process.

Semantic-based search — Deepset

The training also helps the NLP model in predicting the words what is going to come next. It is nothing but training a human child on a song. Once you sing half the poem, the child recites the other half. This is something that we see in multiple applications’ auto-complete features. It also helps in understanding the context. One of the main use cases is when you search for plant and planting equipment on a retail website, and next when you search for clothes, the website might also show you some clothes required for planting purposes. Isn’t that interesting!


Clustering is one of the simplest and highest used AI algorithms. It is mostly used when the input-output combinations are not known. For example, if we do not know the market segmentation and how different types of products, stores, or customer segments exist to be used for organization and marketing purposes. Once the groupings and subgroupings are done, a human can take a look and make sense of why these grouping exists and then tag them accordingly. For example, products can be clustered by type, shape, occasion, materials, features, price, style, design, color, size, family, and brand. And this segregation can hugely help the store managers to organize the store for efficient use and high sales.

Advanced logistic regression

This is another highest used algorithm same as clustering. This does not even involve training. The logistic regression models are fed with input-output sets of data with which the model makes a rule. This rule is used to predict sales and forecast seasonal demands. It also helps in effectively managing the store space and large storage spaces, hence optimizing costs.

So, to summarize AI is a boon for the retail industry since the stakes are low and the benefits are high. There are numerous startups and smaller companies that target bits and pieces and soon there will be companies that aggregate all services and provide to retail industry giants on scale. Or even one step further, the giants can themselves build by acquiring most of them. Regardless it is here to stay and make a large-scale shift instead of just making a dent.



Deepak Singh

Product Enthusiast — Utilizing the power of AI and Design to rethink possibilities and reframe the problem statement! Website: