What is predictive AI?

Predictive artificial intelligence (AI) refers to the use of machine learning to identify patterns in past events and make predictions about future events.

Learning Objectives

After reading this article you will be able to:

  • Define predictive AI
  • Explain the core concepts that make predictive AI possible
  • Contrast predictive AI vs. generative AI

Related Content


Want to keep learning?

Subscribe to theNET, Cloudflare's monthly recap of the Internet's most popular insights!

Refer to Cloudflare's Privacy Policy to learn how we collect and process your personal data.

Copy article link

What is predictive AI?

Predictive artificial intelligence (AI) is a computer program's ability to use statistical analysis to identify patterns, anticipate behaviors, and forecast future events. The field of statistics has long been used to make predictions about the future; predictive AI makes statistical analysis faster and (theoretically) more accurate via machine learning and access to vast amounts of data. While its predictions are not guaranteed to be correct, predictive AI can help businesses prepare for the future and personalize experiences for their customers.

Imagine Joey is a fisherman who needs to know what the weather will be before setting out on his boat. Over the last few months, every time Joey sees a red sky in the morning, he has experienced a storm. Joey begins to conclude that whenever he sees a red sky, he should take it as a warning that a storm is coming. Predictive AI reaches similar conclusions, but by analyzing thousands of factors (instead of just the color of the sky) and decades' worth of data (instead of just several months).

Predictive AI is just one of many capabilities offered by AI, which refers to a set of abilities computers can have that imitate human cognition.

How does predictive AI work?

'Big data'

In statistics more data generally results in more accurate analysis. For example, an opinion poll must have a minimum number of respondents to be considered reliable, and scientific studies need to be repeated several times to be considered statistically significant.

Similarly, predictive AI requires access to vast quantities of data — "big data," as it is often called. The more data provided, the better the predictions. An opinion poll may have a few thousands respondents. A predictive AI model could take into account thousands or millions of opinion polls that have been conducted in the past in order to make predictions about public opinion or upcoming elections.

Machine learning

Machine learning is a subset of AI. It is a method for training a computer program to identify data without human intervention. Given enough examples of user behavior on a website, for example, a machine learning model could learn to sort automated bot traffic from human traffic on the website. Or, given enough photos of a sky and information about weather, a machine learning model could learn to identify a "red" sky and associate certain kinds of skies with stormy weather.

In predictive AI, machine learning is applied to the vast data collections described earlier. A predictive AI model can process huge data sets without human supervision.

Identifying patterns

Just as Joey the fisherman identified the pattern of red skies in the morning being associated with a storm coming, predictive AI learns to associate certain types of data or certain occurrences. Predictive AI can look at hundreds or thousands of factors to identify patterns — which indicate events that can recur in the future.

What are some use cases for predictive AI?

The applications for predictive AI are vast and wide-ranging. Having some idea of what is coming in the future can be a huge advantage for a business, even if such predictions are not always accurate. Some of the possible predictive AI use cases include:

  • Inventory management: Predictive AI can help identify times when consumer demand is likely to be higher and a brand or a store should have more items in stock.
  • Supply chain management: Similarly, predictive AI can help determine when times of road congestion will occur or when more trucks are needed to meet spikes in user demand.
  • Personalized user experiences: Predictive AI can help anticipate user behavior based on past activity.
  • Healthcare: Given enough data, predictive AI could help forecast potential future health conditions based on a person's medical history. (Health data, of course, is subject to strict regulatory frameworks like HIPAA.)
  • Marketing campaigns: Just as predictive AI can anticipate user or customer behavior, it can help prognosticate what kinds of content or products that prospective customers may be interested in.

Predictive AI vs. generative AI

Predictive and generative AI both use machine learning, combined with access to lots of data, in order to produce their outputs. However, predictive AI uses machine learning to extrapolate the future. Generative AI uses machine learning to create content. The predictive-AI version of Joey the fisherman tells his fellow fishers when a storm is coming. The generative-AI version of Joey writes a novel that imagines various interactions between weather and fishing voyages.

In a sense, generative AI is similar to predictive AI, as it uses statistical analysis to "predict" which words and concepts belong together. But the goals for generative and predictive AI are different, the machine learning models they use are different, and the use cases are different.

How does predictive AI use embeddings?

Like most types of AI, predictive AI requires the ability to query databases quickly and efficiently, and to find relationships between similar items of data. A database of embeddings makes similarity queries possible.

Embeddings are a way to store information in a form that allows for the identification of similarities and relationships. Created by unsupervised neural network layers, embeddings turn items of information into vectors, placing them within a mathematical space in relation to the other items of information in the data set.

Embeddings that end up clustered together can be considered relevant to each other, and this allows for rapid pattern identification. If "red sky", "storm clouds", and "rough weather" are all close to each other in an embeddings database, a predictive AI model can begin to identify when a storm is coming.

Read more about AI.