Two robots insulting each other: is this the future of procurement automation?
Start Trek’s android hero, Data, is engaged in a quest to understand and adopt all the aspects of being human. Why then, is he so polite and even-handed? Where’s his tendency to abuse strangers in social media?
Recent developments in Artificial Intelligence (or a small subset thereof) trying to mirror human behaviour, have seen that mirror reflect some not too pleasant aspects of our nature.
Chatbots, it seems, are finally coming of age. Originating from the early days of AI as potential tools to pass the Turing test, chatbots, whilst still lacking the convincing conversational skills of Commander Data, are suddenly coming amongst us as marketing tools.
Microsoft, messaging app Kik and shortly Facebook are all making chatbots available in their business app stores to allow marketeers to use them for engaging with the audience that inhabit the social media and chat spaces.
It is somewhat surreal that the proliferation of real-time on-line chat agents in the support areas of companies’ web sites will make this a fairly natural transition that may well go unnoticed.
Let me explain. Whether you need to find out about a service or have a technical enquiry, its standard practice, pretty much, to be able to click on a “live chat” button and within a few seconds be “talking to” [insert randomly selected name here].
What quickly becomes apparent on many occasions is that there is a degree of language barrier involved. Of course, the support agents could be anywhere on the planet and maybe working from their own homes, so the dialog generally has to involve some fairly specific and detailed exposition on behalf of the customer to ensure a cogent and coherent response from the support agent. If the customer provides the full spread of detail and data up front, the agent can assess and arrive at an appropriate response.
But this is exactly what we’ve had to do so far with chatbots. To avoid repeated responses such as “can you be more specific” and “do go on” we need to be very clear and logical to be understood. So, as the bots become more sophisticated we may never notice the difference between a bot and a non-native speaker. That may well mean that the on-line agents will be replaced pretty quickly. One bot could conduct many thousands of concurrent conversations. Imagine the efficiency boost!
Imagine the customer frustration if it’s not perfect!
Microsoft’s recent mishap with their chatbot, Tay, demonstrates a darker side to our interaction with technology. When we know it’s a machine we’re dealing with we (present company excluded) have a tendency to try and subvert it, to catch it out, or in Tay’s case to turn it into a foul-mouthed ranting xenophobe.
Then again, we all have direct experience of humans treating each other with equal lack of respect in online forums. It even has a name, Godwin’s Law. It goes like this: “As an online discussion grows longer, the probability of a comparison involving Nazis or Hitler approaches 1” We humans, it seems can’t help descending into abuse when there are no apparent consequences and in trying to replicate human behaviour it seems that Tay has been just a little too successful.
Perhaps, then, we might set a bot to catch a bot. The same chatbots are being made available to us as individuals and consumers and in the growing quest for increasing automation of procurement processes we may see the self-same bots becoming a key part of the dialog between buyer and supplier.
But when the buyer-bot and supplier-bot are making offer and counter-offer in internet time, and cursing each other with ever-escalating ferocity, will that mean the effective end of negotiation?
Technically it could, but then, who would want that? Neither buyer, nor supplier, I’d wager. Automation should be about accelerating drudgery out of our processes, not replacing those parts where human intelligence adds maximum value.
Thus, in the search for Turing-approved AI we might need to assume that the chatbots are, in fact, foreigners who speak a different language and we should take great care when asking questions and interpreting answers before making decisions. But that’s what procurement professionals do day-in, day-out.
Procurement, then, may have a thing or two to teach the technologists of today about how business of tomorrow should be conducted. The bots may carry out many of the transactional processes but it will still take human smarts to drive value into the business.