There was a joyful moment in our house today when a phone call via an automated system to resolve a bank card failure led to a conversation with a helpful human being (just press * in this instance) who resolved the problem immediately.
(I know, don’t I usually start with a whinge?)
The moment of delight occurred because the automated system could not cope with the delay and deviation invoked by its demands for data that could not be provided in the response time available and so it offered a conversation instead.
The call highlighted the limitation of the system algorithm or, more fairly, the limitation of the thinking of the people who designed the algorithm. Let’s not blame the machine, it is only doing what it was programmed to do.
And here is the issue, the problems and challenges we face that require us to deal with such algorithm driven problem solving devices usually do not quite fit the assumptions (we will call them that instead of biases and prejudices) either explicit or tacit of the designers of the system. The aim of the automated system is to make it cheap for the organisation to respond to enquiries and requests and convenient for the customers. So far so sensible; but not if the designer pushes too far and the variety of the system is unequal to the variety of the enquiries the customer presents. As Peter Dudley and I have said before, no matter how long you make the process chart you never quite reach the customer!
When the system doesn’t solve the problem the cost of failure is commonly displaced to the customer so the immediate cost to the provider is £0! (As I write this, I have just waited 24 minutes for a call not to be answered – but I can’t invoice the company for my time!) In the longer term the cost bites the business because (assuming an alternative provider exists) the customer takes their custom elsewhere. The zero cost and customer loss events are displaced in function, time and space so the business does not make the connection, it does not recognise the pattern.
US quality guru Philip Crosby said, at least as far back as 1980, “Quality is Free” by which (for the purposes of this blog) we shall take him to mean that the cost of doing things right first time is always lower than the cost of putting things right!
Despite the extravagant and grandiose claims of the information industry, the current state of many ‘AI’ systems (voice response, chatbots, automated decision devices) is infancy. They are relatively simple systems capable of dealing with relatively simple problems – particularly when compared with the sophistication of customer needs. Primarily through various approaches to pattern matching (i.e. “this looks like that”) such systems do not originate or ideate but imitate.
If the problem presented is not close enough to the established pattern it will not be recognised. The range of questions and queries that customers need to ask are commonly like, but not identical to, those designed in to the system. They are often more complex or nuanced than the capability of the system to address them, they need people as interpreters. People are good at variety, nuance and approximation, even at working out the question you SHOULD be asking as opposed to the one you did!
Accepting the limitations of the automated systems may mean constraining the range of questions and responses to which its algorithm(s) are adapted, the rest being dealt with by human beings. On average, each query might be a little more expensive to deal with, but if all queries might then be resolved at the first attempt that saves the customer time and money and makes them happy!
It compensates the company for not being cheap by saving time and money in two ways:
retained customers are inherently more profitable than new ones in most instances (because you have already absorbed the cost of acquiring them);
reducing the waste of time and money dealing with the second, third, fourth, fifth attempt to solve the same problem.
Design your system to:
deliver a first-time fix;
utilise appropriate technology appropriately (and don’t buy the vapourware);
utilise appropriate people appropriately;
optimise the balance of the two.
Don’t do it cheap, do it right!