Could a machine take your job?

When we speak about the possibility of robots taking our jobs, and the idea of machines working more efficiently than us, historically we focused on manual labour.  That is to say we spoke about factory jobs, production lines, things where we as people carried out repetitive tasks with little variation.  These roles were relatively easy to replace with machines.  We have now lived with an automotive industry that for decades has used automated assembly lines.

There was an argument at the time that these steps forward would mean the loss of jobs for people, and that was true.  There was also an argument that said the machines would still need to be maintained and while jobs would be lost there would also be jobs created to maintain the automated assembly lines, and again this was true.  Ever since that time however there have been competing views, those who hold on to the first, and those who hold on to the second.  In my view I have considered both points of view and I see the first as valid, the second as not.  The reason I see it this way is because I have studied Artificial Intelligence, and developed Agents, and one thing remains true throughout everything I studied - if you can teach it, they can learn it.  Almost everything you do as a human being you were taught how to do.

The barrier to machines taking the majority of people's jobs remains the same as it did back then - teaching.  Or to be more specific, machine learning.  It wasn't possible at the time for a machine to maintain itself and to accurately diagnose the wealth of problems that could go wrong many of which would be unforseen, and in that vein they could not be programmed to respond to the unforeseen.  Today however this isn't true.  Advances in machine learning and the way we program have allowed us to create Agents which can adapt over time.  This barrier is however crumbling, and as it crumbles it is being replaced by other more traditional barriers, such as cost, resources, and politics.

In London the Docklands Light Railway [DLR] has been operating since 1987.  The DLR is a driverless train system.  While some staff are still required primarily for security and in the event of failure, the trains for the most part run on their own.  In more recent times we have seen driverless cars enter into the mainstream.  The question of job security is once again coming to the forefront, and the question of whether a machine could one day take your job is something many people now ask themselves.  You still get people who defend the viewpoint that their jobs could never be replaced.  You get people who think that their job requires a human touch, and that a machine could never learn the intricacies of their work.  You still get people that work in places like call centres and think that people won't talk to machines they want to talk to people.

People won't want to talk to machines - are you sure about that?  I would argue the only reason that was ever the case was the lack of engagement machines provided in the past.  I would argue it had nothing to do with the fact it was a machine and would go so far as to say the fact that people so openly embrace Siri, Cortana, Alexa, and OK Google so freely and are perfectly comfortable talking to their machines that this really is not the case.  The question then turns to whether a machine can learn how to do your job.  Well as far as call centres are concerned the answer to that is yes as far as I am concerned and the reason why is one basic fact:

Basic Fact #1 about call centres:  All calls are recorded.

You're used to being told "Your call may be recorded for training and monitoring purposes" but what you probably don't know unless you have worked in a call centre is long before you hear that message, your call is already being recorded.  I have worked in a call centre and I can tell you quite simply, every single call is recorded, and we can hear you before you can hear us.

Every call is recorded for quality assurance, for legal disputes, and yes for training and monitoring purposes.  The call centre I worked in dealt with financial services and there were around 200 people on the floor.  Each one handles about 50 calls per hour, we can say 40 for lenience sake and imagine the majority never met their KPI targets.  On the floor there would be around 200 staff at any given time, working on rotation some would leave and others would replace them at the end of their shifts.  The centre handled calls from 8am to 8pm.  They handled calls 52 weeks of the year with reduced service on 28 days of holiday which included bank holidays. 

With 52 weeks in the year, each 5 days long Monday to Friday, that is 260 days, minus 28 for holidays makes it 232 days worked per person on average per year.  Let's drop it to say 220 to allow for sickness and absence.  So we have 220 days worth of calls per call handler, that's for one year.  For 12 hours per day, 40 calls per hour, 200 staff on the floor, 220 days of the year, that makes 21.1 million calls.  The centre keeps all calls for 6 years in accordance with the financial regulations so they have about 126.7 million calls on file.  If you say each call lasts about 1 minute, then divide by 60, they have 2.1 million hours of phone calls on file.

With 2.1 million hours of recording, a machine learning algorithm could digest the lot and learn very quickly how to do your job.  There is no practical barrier to this technologically today, the only barrier to this is now cost.

Could a machine take your job?  Yes quite easily.