Transport and logistics giant UPS Inc. is turning to artificial intelligence to better communicate with its customers, beginning with a "chatbot"—or chat robot—platform it launched on Monday.

Customers who have package-tracking questions or want to find a nearby UPS Store can send a text message to the chatbot, and its computer algorithms will recognize keywords and generate an appropriate response, UPS said. Users can access the tool through the Facebook Messenger, Skype, or Amazon platforms, which means that fans of Amazon's voice-operated "Alexa" application can hold spoken conversations with the UPS chatbot instead of typing.

Atlanta-based UPS currently fields questions from customers through its mobile app, website, toll free phone number, email, and "LiveChat," options, all of which rely on human operators to field questions. However, the company said it does not plan to use the automated chatbot to replace those workers.

"We don't see it as replacing our call center; this is a new channel to answer questions that might not have been answered before," said Derek Banta, UPS' director of digital channel and mobile applications, in a phone interview.

For example, multi-tasking shoppers buying holiday gifts on their lunch break could use the chatbot to ask about shipping rates and find the business hours of the nearest UPS Store, he said. In future applications, UPS plans to apply the chatbot technology to additional platforms, including the UPS My Choice product, which allows customers to manage the delivery time and location of incoming packages.

"We are excited to get into the chatbot world, but this is just a first step in our bigger goal of leveraging [artificial intelligence] or machine learning to meet customers where they are," Banta said. "It's important to meet customers on their own terms. So if you're on Facebook, you can now use Facebook Messenger to answer questions instead of having to leave to go to UPS.com."

Using machine learning techniques, the UPS chatbot will get better over time, refining its ability to identify human grammar and replying with more accurate answers, he said.

"We can refine the AI using conversational learning; think of it like learning a language," said Banta. "The machine has to learn that 'package' and 'shipment' are one and the same, so if you ask it 'track a package' or 'track a shipment,' it means the same thing."

As it continues to train the tool, UPS is wary of some recent missteps with the technology, such as a failed effort earlier this year by software giant Microsoft Corp. The Redmond, Wash.-based company unveiled a chatbot named "Tay" intended to field customers' questions. But Microsoft was forced to cancel the project just days later when the system's machine learning process had explored the darker corners of the Internet and began replying to customers with racist and sexist rants.

UPS plans to avoid that scenario by assigning human supervisors to monitor the machine's progress, Banta said.