So you’re thinking about incorporating AI into your insights process but you’re not sure which is the most useful to you. The latest developments in AI are all talking about deep learning and neural nets. But do you need a neural net to harness the power of AI and benefit your business? Most likely not. I’m going to run through some typical insights needs and which type of AI can help and how.
Based on previous data, you can build AI models that will make predictions using regression models. A typical example is a price predictor for stocks, or other goods and commodities. However less typically quantitative trends can also be predicted, such as what color is likely to be on trend next season assuming that there is relevant data available.
When we think of classification models some of the most well-known examples are language detection and image identification.
Sentiment analysis is a type of classification that can be useful to insights managers to quickly flag or identify positive or negative sentiment in reviews, or other textual data.
A custom trained machine learning model can be beneficial for specific industries. For example if your work primarily deals with mining for insights in the car industry, or luxury travel, or alcoholic beverages, a custom model will be able to take into account specific phrases and vocabulary that may not be typically found in general conversations.
Some machine learning models specialize in spotting patterns in data. This can be useful in discovering previously unnoticed trends, patterns or groups that are present in data, or simply to lend a human researcher an extra ‘pair of eyes’ so to speak.
Examples of uses for this type of AI in insights include basket analysis to see what shoppers tend to purchase together, and creating customer personas by seeing what characteristics define a consumer typology.
Models that identify ‘themes’ in text data also do so by looking for patterns. This is a more intricate type of text processing that can benefit from custom processing of the data and model set-up.
The latest and most sophisticated developments in machine learning and AI — deep learning, recurrent neural networks, reinforcement learning are to create models that can make decisions similar to how a human would do it. A good example is self-driving cars and robots like these, and even bots that can decide on what words to use to reply to you in a chat box.
Creative AI is cool and funny and we love to laugh at the weird picture a computer created or the not-quite poem it made. There are ways however that machine ‘creativity’ can but useful in a professional setting primarily as an augmentation to human creativity.
Some companies like textio.com are using large databases of text to analyze tone and give recommendations and prompts for crafting messages to a target audience.
In a nutshell, AI is very powerful and can even be a lot of fun. It does however still take a lot of human input to get the best, most useful results from it. The way I like to look at it is that AI’s not here to take over the world, but a good friend and strong thinking partner.
Stef Lai is the founder of In Place, a qualitative consultancy that brings you better insights with the help of machine learning, and demystify AI at the same time.
Vote for In Place in the IIex start-up competition!