During his research into the ruthless side of the modern economy, the journalist Simon Head discovered a fascinating paper, published in 1997, about a customer relationship management IT system in action.
It was written by two researchers at Xerox’s Palo Alto Research Centre in Silicon Valley, and they describe a corporation called MMR, a thinly disguised version of Xerox, and their IT system called CasePoint, which was designed to automate the conversations between customers and agents. The idea was to cut the cost of sending technicians out to repair Xerox machines. It would all be done automatically by call centre staff over the phone.
CasePoint didn’t work. Call centre staff were supposed to take down exactly what the customers said and the system would feed it to the experts to come up with solutions. The call centre staff were only allowed to ask questions exactly as they were worded on the screen.
But what really amazed me about the story was what he said:
Because, after all, as C. S. Lewis put it, in the mouth of Mr Beaver: "Take my advice. When you meet anything that's going to be human and isn't yet, or used to be human and isn't now, or ought to be human and isn't, you keep your eyes on it and feel for your hatchet."
CasePoint didn’t work. Call centre staff were supposed to take down exactly what the customers said and the system would feed it to the experts to come up with solutions. The call centre staff were only allowed to ask questions exactly as they were worded on the screen.
The trouble was, when it came to the point, the customers used ‘unauthorised’ language of their own which the system couldn’t understand. By the end of the trial period, the researchers couldn’t find one example where CasePoint had done its job properly. That is what tends to happen when human beings are excluded from systems.
The trouble was that, when the managers and software designers were told this, they decided it was an irrelevant detail and that they should carry on with the system regardless. They wanted to see how far they would rely “exclusively on machine expertise as a substitute for agent knowledge”. The answer was: not very far.
This is the dream of IT consultants: a machine that can have a conversation with a human being without them knowing they are talking to a computer. In fact, most people who have been phoned by a computer will tell you that, the more like a human being the machine is, the more unnerving the experience.
The trouble was that, when the managers and software designers were told this, they decided it was an irrelevant detail and that they should carry on with the system regardless. They wanted to see how far they would rely “exclusively on machine expertise as a substitute for agent knowledge”. The answer was: not very far.
This is the dream of IT consultants: a machine that can have a conversation with a human being without them knowing they are talking to a computer. In fact, most people who have been phoned by a computer will tell you that, the more like a human being the machine is, the more unnerving the experience.
Computers show little signs that they will ever be able to confront a human personality with their own. Maybe they will one day, but I doubt it: all they can do is to fake a personality, usually with some marketing or processing intent. Until they go beyond this into genuine artificial intelligence, their failure to do so can only be hugely expensive.
CasePoint is an extreme example of what happens when IT takes over functions it shouldn’t, where it deliberately excludes closer human contact. More about this in my book The Human Element.
CasePoint is an extreme example of what happens when IT takes over functions it shouldn’t, where it deliberately excludes closer human contact. More about this in my book The Human Element.
But it is also an example of the great unchallenged nonsense of a 'shared back office' service. The key assumption is that customer-facing systems can be automated entirely - what the government calls 'digital by default'. It also assumes that untrained system operators will just service requests, find a place for each query on the software, and hand it back to automated back office systems - or, failing that, experts.
There is very little evidence that this idea, embraced by government and consultants alike, saves money. Of course automation often seems to save money, but consultants very rarely subtract - so they fail to see the diseconomies of scale until it is too late.
Only when the costs mount, as the systems thinker John Seddon explains, does anyone wonder (and often they don't even then) whether the automation has made the system frustrating and inflexible, and therefore wasteful - because some people have to come back again and again (what Seddon calls 'failure demand').
And the most obvious times when we encounter this phenomenon is the 'Press button 2' syndrome. Which is why the story of the retired IT manager timing each stage through different call centres is so fascinating - six minutes for each of the four levels at HM Revenue and Customs.
But what really amazed me about the story was what he said:
"In an ideal world, he said, companies should just offer different phone numbers for different services."Isn't that what we had only a few years ago? There is a good reason why we don't - because people's queries often don't fit the boundaries set down (as we find so often pushing buttons).
I even remember Simon Hughes' campaign for the London mayor proposed 'One Number for London' as its main policy demand - about the least exciting political demand I have ever come across. If he had won, and the One Number was up and running - how many different options would people have to go through now?
No, it only works if there is a human being, sensible, flexible and empowered enough to make things happen, at the other end. That requires IT support at their end, not push buttons at our end.