Pi, released this week by the firm Post Intelligence, employs a neural network to create a unique model based on each user’s habits.

Today, a newly renamed firm is releasing a social media marketing tool that uses artificial intelligence to increase interaction and, in some cases, even compose the post itself.

In 2015, Bindu Reddy and Arvind Sundararajan, both formerly of Google, started Post Intelligence in San Francisco under the name MyLikes. The renaming comes as the beta version of the company’s Pi intelligent social marketing software for Web and Android is released.

In order to forecast engagement (such as likes and shares), propose material, learn what works, and even produce posts, Pi monitors social trends and a user’s history of social postings.

According to the firm, unique models of engagement prediction are constructed by training a deep neural network on each user’s social media feed. Predicted levels of participation range from 0 (very little) to 10 (very engaged) points. Reddy assured me that with an 80% confidence level, Pi can correctly forecast whether engagement will be low, average, or high.

With Pi, a user may schedule posts across Facebook, Twitter, and Pinterest and publish them at certain times. We’re currently working on adding Instagram and Snapchat support.

According to Post Intelligence, the model can really design and deliver a brief post on a specific topic after learning the user’s posting tendencies over time. The following is an example of how a political-minded user of Post Intelligence may express herself:

“Hi Pi, Share my dismay at the new healthcare plan and the loss of insurance to millions of Americans by tweeting this. The AI would deduce the meaning of the words, quickly locate information about the healthcare bill, and write a tweet with a link to an article on the subject.

Pi also features ad support similar to a product released back when the firm was known as MyLikes. Based on similarities to the influencer’s most successful posts, it will suggest more articles from over 30 participating publications to those with over 10,000 followers on the supported platforms.

The recommended material may be shared by the influencer as a link with a blurb and image or video from sources like Huffington Post, Bored Panda, Simplemost, and others. The influencer earns a commission when their followers use the link or watch the video.