As The Wall Street Journal reported in March, average call duration at Allstate Insurance call centres dropped to 4.2 minutes from 4.6 minutes after the company deployed Amelia.

Her maker, Chetan Dube, left his professorship at New York University to start IPsoft, an automation software company, convinced that using remote workers in India would be nowhere near as efficient as replacing the U.S. and European workers with cloned human intelligence.

Amelia and her kind are not enhancers of labour productivity — like faster laptops or better database systems. They are designed to replace workers. They are not as good as human workers, but they are significantly cheaper.

Until recently, most white-collar, service-sector and professional jobs were shielded from automation by humans’ cognitive monopoly. Computers couldn’t think, so jobs that required any type of thinking — nuclear physics professor, florist and everything in between — required a human. But a form of artificial intelligence called “machine learning” has given computers skills like reading, writing, speaking and recognizing subtle patterns. Thinking computers like Amelia are opening a new phase of automation and bringing the pluses and minuses to a whole new class of workers — those who work in offices rather than farms and factories. These people are largely unprepared.

U.S. workers today also face direct wage competition from highly skilled, low-cost foreign workers working (virtually) in U.S. offices. Web platforms like Upwork — essentially eBay for free-lancing — allow employers to source talent from across the globe, often at a much lower cost. Think of this as telecommuting gone global. Think of it as telemigration. The combination of this new form of globalization with this new form of robotics — call it “globotics” — is really something new. The most obvious difference is that it’s affecting people working in the service sector instead of the manufacturing and agricultural sectors — about 129 million Americans as of last month, according to the U.S. Bureau of Labor Statistics.

Amelia is used by over 20 of the world’s leading banks, insurers, telecom providers, media companies and health care firms. Research had shown that customer satisfaction with phone-in helplines is directly tied to empathy shown by the agent handling the call, so Amelia’s maker added a psychological module to her algorithm. She is aware of the emotional state of the person with whom she is speaking, and she adapts her responses, facial expressions and gestures to better communicate. Amelia can’t handle everything. Sometimes she passes on information to her human colleagues so they can take over. But Amelia is curious. The software hangs on the line listening to the humans — and the resolution of a problem. She then adds these new tricks to her knowledge-management system. Once her learnings are approved by her human supervisor, she can answer similar queries herself in the future.

In the past, some AI “advances” were a flash in the pan. But many companies have introduced Amelia-like software. Bank of America rolled out Erica in the summer of 2018. She can let you know when your checking account has dipped into the red or suggest putting additional funds toward your Visa, based on your monthly spending. JPMorgan’s white-collar robot is called Contract Intelligence. Capital One has Eno. IBM is selling many Amelia-like virtual assistants under the brand Watson. Salesforce offers Einstein, SAP has HANA and Infosys has Nia.

The sophistication of AI systems like Amelia, Watson and Erica will displace many service-sector jobs — but they will eliminate only a few occupations. This is what automation has always done. Tractors, for example, automated some farm chores, but they did not eliminate farming as an occupation. We just needed fewer farmers. This is what we’ll see all across the service sector in the coming years.

Many studies have tried to estimate the total impact of recent, AI-linked automation on jobs. A 2016 study by the consulting firm Forrester suggests that 16 percent of all U.S. jobs will be displaced by automation in the next 10 years, with white-collar professions expected to be the hardest hit. About half of the job destruction will be matched by job creation — positions such as “robot monitoring professionals,” data scientists, automation specialists and content curators — resulting in a net loss of 7 percent of jobs by 2025.