Why this start-up is teaching A.I. chatbots to have 'manners'

Key Points
  • Automated applications powered by artificial intelligence, also known as bots, are still "dumb," "rudimentary" and "frustrating," says a former Facebook product exec.
  • Intercom, a start-up that sells bot-based messaging apps, says it has taught them to have "manners."
  • The announcement comes after Facebook said in June it taught its bots to "negotiate."
Mark Zuckerberg introduces the Messenger platform at the F8 summit.
Josh Edelson | Getty Images

The technology industry is infatuated with automated software programs powered by artificial intelligence.

YouTube is using bots to recommend music videos, 's are recommending online purchases and 's are making it easier for businesses to communicate with users of its Messenger app.

The consulting firm Accenture even said in a report that the technology "will double economic growth rates in 12 developed countries and boost labor productivity by up to 40 percent" by the year 2035.

There's just one problem: "Bots are dumb and rudimentary," said Paul Adams, vice president of product at Intercom, a San Francisco-based start-up that sells bot-powered messaging services to businesses.

This week, Intercom is trying to do something about the shortcomings of such automated software.

On Wednesday, the company is unveiling what it calls Operator, an update to its bot operating system that Intercom claims can do what no other bot technology can: teach bots to be polite.

"Today's bots don't have manners," said Adams, a former product manager and head of global brand design at Facebook. "They don't take 'no' for an answer, they keep spamming and are frustrating" for their users, he said in a phone interview.

Today, Intercom bots handle simple, repeatable tasks such as collecting contact information, setting response time expectations and automatically answering questions by sharing relevant articles.

Operator will make sure Intercom's bots are more polite by enforcing the following rules:

  • Don't interrupt when a customer is typing.
  • Don't suggest articles if a customer has tried to self-serve.
  • Don't spam customers or leads with more messages if they don't engage initially.
  • Withdraw from a customer conversation after a human gets involved.

The software will sit in the background of its various bot products for functions like sales, marketing and support.

The innovation comes 18 months after Microsoft was forced to end an experiment with a bot called "Tay" that was led into making racist statements by users, and just weeks after Facebook caused confusion with a paper about how it taught bots to negotiate — some reports concluded incorrectly that the researchers shut the experiment down because the bots were communicating in their own made-up language.

Founded in 2011, Intercom has 350 employees, 20,000 customers and has raised $116 million in funding, according to its website.