How TGI Fridays Generated $150 Million in Revenue Using Out-of-the-Box AI

At the Marketing AI Institute, we read dozens of articles on artificial intelligence every week to uncover the most valuable ones for our subscribers (become a subscriber today), and we curate them for you here. We call it 3 Links in 3 Minutes. Enjoy!

In the last year, restaurant TGI Fridays has doubled their off-premise sales, a total of $150 million in revenue, just using artificial intelligence. Instead of spending thousands of dollars designing a unique AI system specifically for TGI Fridays, chief experience officer Sherif Mityas outsourced to three main software tools. VentureBeat has the lowdown on these companies.

Amperity, a nine-month-old program, is used to stitch all of TGI Fridays’ data together. It takes email engagement, in-store receipts, loyalty program information, app usage, and more into consideration. Then, using machine learning techniques such as decision-tree branching, it sends personalized offers to users.

Chatbot startup Conversable is used by TGI Fridays to support all their messaging channels, such as Facebook, Twitter, Amazon, and Alexa. Using natural language processing (NLP) to understand what the customer is asking and generates a new answer every time versus responding with options A, B, and C, like most chatbots.

Lastly, TGI Fridays partnered with Hypergiant to create a Virtual Bartender named Flanagan. So far, it has created over 300 different taste profiles based on bar customers’ unique moods, taste, and past behaviors.

AI Beats Top 1% in Dota 2 Video Game

AI hit a new milestone this week: it beat the top one percent of human amateurs at the game Dota 2. The contest was performed at OpenAI, a research lab founded by Elon Musk and Sam Altman, reports The Verge.

Last August, a similar, but not as impressive feat was accomplished. OpenAI introduced a system that could beat the best players 1v1. Now, OpenAI has upgraded its bots to play 5v5, which requires more long-term planning and coordination. At any given time, the AI bots have to decide between 1,000 possible moves while taking in 20,00 data points that affect that move.

OpenAI’s motivation behind teaching AI bots to play video games is to use the problem-solving skills developed to solve real-world problems that resemble video games, such as managing a city’s transportation system.

“This an exciting milestone, and it’s really because it’s about transitioning to real-life applications,” OpenAI’s co-founder and CTO Greg Brockman told The Verge. “If you’ve got a simulation [of a problem] and you can run it large enough scale, there’s no barrier to what you can do with this.”

How to Assess Machine Learning Risk Within Your Company

Many conversations have been started on the ethical and legal framework needed to manage machine learning. This week, a paper was released by the Future of Privacy Forum and startup Immuta on the steps your business can take to assess the risk of using machine learning within your organization, according to Fast Company.

The paper, modeled after a 2011 Federal Reserve document, recommends setting up three lines of defense for handling AI risk.

The first line suggested is data scientists that define exact assumptions and goals around a project. This should be followed by a team of legal experts that review assumptions, methods, documentation, and data quality. The final line of defense should involve reviews of the overall assumptions around the model and how they’re working out.

Every organizations’ method will be different but the authors hope this will serve as a starting point for assessing artificial intelligence risk. They also expect to continue making updates to the report, as this is their first attempt at outlining such a plan.