The rise of government "welfare robots" and AI to help decide if Universal Credit claims are true have been detailed in a new investigation.
The Department for Work and Pensions (DWP) is using 16 'bots' to perform human tasks, like sending automated messages about applications and payment.
Separately, it's understood the DWP is also testing machine learning to comb through Universal Credit applications to work out which need further scrutiny.
It is thought this artificial intelligence works by comparing claims to thousands of previous applications - then flagging whether they are likely to be able to be handled swiftly, or are complex and need further investigation by a human being.
DWP chiefs insist this process speeds up the time to handle claims for childcare and housing costs under Universal Credit - and reduces the amount of evidence needed from claimants.
Officials also stress any final decisions are made by a human being.
But the rise of automation generally has raised fresh fears that vulnerable people could fall through the cracks.
The six-in-one benefit has had to have billions of pounds worth of improvements after the benefit - bundled up with welfare cuts and made digital-only - left desperate people short on rent and turning to foodbanks.
Lib Dem MP Heidi Allen branded the findings "shameful", adding: "Confirms my worst fears - that the Government think it’s okay to treat people as data, not human beings.
"The DWP needs to be pouring desperately needed compassion into the welfare system, not bots."
Many of the DWP's robotics and AI developments come from a so-called 'Intelligent Automation Garage' which was set up in 2017 and has cost more than £7m.
The DWP boasted that such 'intelligent automation' saved 42,000 human hours in the processing of fitness-for-work notes, and 22 minutes per bundle on PIP appeals.
Part of the work involves using private New York-based firm UiPath, whose chief executive Daniel Dines said last month: "I want a robot for every person".
UiPath boasts its work has already saved millions of pounds each year, including using 12 robots to clear a backlog of more than 30,000 pension claims in two weeks.
And the firm previously said the DWP had a target to double from 10 to 20 automated processes by March this year.
Frank Field, chairman of the Commons Work and Pensions Committee, told The Guardian, which ran the investigation: "We've already seen in the gig economy how workers are managed and sacked, not by people, but by algorithms. Now the welfare state looks set to follow suit."
Dr Lina Dencik, of the Data Justice Lab at Cardiff University, added: "There are concerns that the government is accelerating the automation of the welfare system without a proper evidence-based consultation about its impacts."
Statements by some of the DWP's top staff have suggested automation is set to rise further.
A 2018 blog by DWP senior product owner Shaun Williamson said officials were "exploring the potential of chatbots" - claiming they could cut calls about sickness benefit by 200,000 per week.
A DWP spokeswoman said there are not currently any claimant-facing chatbots in the system.
The DWP said it understood some people need help with technology, so there are staff in Jobcentress to help people access computers, and £39m of spending through Citizens Advice to support people who are hard to reach.
A DWP spokeswoman said: “We are committed to having a compassionate safety net, on which we spend £95 billion, and creating a digital service that suits the way most people use technology.
“Automation means we are improving accuracy, speeding up our service and freeing up colleagues’ time so they can support the people who need it most.”
