A Starting Point for Building AI

Five Ways to Think About Integrating Artificial Intelligence into Product Development

The Challenge

Market leaders across the United States understand that Artificial Intelligence (AI) has the power to disrupt their businesses, and they are all attempting to integrate AI into their operations with urgency to remain competitive. This might be a lesson learned from the mobile era when many market leaders were dethroned by startups using new technology faster and better than they did. However, determining exactly how AI has the potential to improve your business can be challenging.

Our Approach

In the Lenfest Local Lab, our mission is to develop experimental products for journalism — another business being disrupted by digital. Since the application of AI in journalism is still in its very early days, we’ve been thinking of ways our experiments can shed light on how AI can be beneficial for news. After attending the 2019 Computation + Journalism Symposium at the University of Miami with the lab’s director Sarah Schmalbach, and researching popular AI startups, we’ve learned more about how other newsrooms, and investors, are approaching this new technology.

The Starting Point

For example, Andrew Ng, a Stanford University adjunct professor and co-founder of Google Brain, claims that AI projects within organizations tend to fall into one of two buckets: New Categories and Improvements in Current Businesses Processes.

This categorization provided a good start for a mental model for thinking about possible experiments, but I believed further segmentation was possible. I started researching various AI products and startups, including the most recent batch of AI startups that came out of Y Combinator, a well-known Silicon Valley startup accelerator. I found that the majority of the projects and companies fit into five categories. I’ve listed them below along with definitions and examples of each.

A diagram of AI product categories.

I. New Categories

Expensive Task Completion — Things that people would like to do, but can’t always because it requires lots of time and money.

Examples:

  • Song identification, such as Shazam’s “fingerprint” technology that allows anyone to identify the song they’re hearing.
  • Video transcription. The AP uses an automated video transcription service that has saved video editors “hundreds and hundreds of hours” of work according to the AP’s Director of News Partnerships, Lisa Gibbs.

Personalization / Categorization — The grouping of content.

Examples:

  • Next article recommendations: AI products like Contextly automatically recommend articles that an algorithm predicts you’ll want to read next.
  • Dog breed classification: AI categorization is being used to determine the breed of a dog by taking and examining a photo.

II. Current Business Processes

Repetitive Task Automation — Things that are currently done that are repetitive.

Examples:

Quality Monitoring or Improvement — Things that require people to observe and make sure something is happening or not (and potentially fix).

Examples:

  • Shoplifter detection: Detecting shoplifters using security cameras in stores, such as Vaak, a Japanese startup that claims to identify shoplifters before they steal.
  • Agricultural produce quality monitoring: Farmers use cameras and conveyor belts, instead of manual labor, to identify and/or remove produce that has gone bad.

Estimation / Probabilistic Decision Making — Things that require accuracy and are already being calculated.

Examples:

  • E-Commerce fraud detection: Companies like Forter are using various data signals (ex. time of purchase, IP address, etc.) to determine which online purchases are fraudulent.
  • Fake news detection: AI can be used to determine a photograph’s authenticity and if it’s being used out of context. Tools such as TinEye’s reverse image search are used for this purpose.

Just the beginning…

Through our research and conversations, we found that there are many ways to categorize AI products. We hope these five can be a starting point for people in journalism and other industries when thinking about how they can integrate AI into their businesses. As we move forward, these categories will guide our thinking but are bound to change.

If you have suggestions, additions, or feedback please reach out via Twitter @lenfestlab or @ajayjapan or leave us a comment.

Ajay Chainani is currently a software engineer at the Lenfest Local Lab. He has prior experience helping build award-winning startups and worked in an advisory role for dozens of other startups in Japan and the United States.


The Lenfest Local Lab is a multidisciplinary product and user experience innovation team located in Philadelphia supported by The Lenfest Institute for Journalism.

The Lenfest Institute for Journalism is a non-profit organization whose mission is to develop and support sustainable business models for great local journalism. The Institute was founded in 2016 by entrepreneur H.F. (Gerry) Lenfest with the goal of helping transform the news industry in the digital age to ensure high-quality local journalism remains a cornerstone of democracy.The Lenfest Local Lab

A Philadelphia-based news product innovation team within the Lenfest Institute for Journalism experimenting openly with ways to reinvent the daily user experience for local news

Thanks to Sarah Schmalbach. 

Leave a Reply

Your email address will not be published. Required fields are marked *