From Mobile World Congress 2016 to the recent F8 ten year roadmap speech, AI is definitely one of the hottest technology trends. And specifically, AI in the customer experience, which  is the front-line of any expectation towards a company has been buzzing for a while as an innovation topic.

News from the several AI experiments are not much reassuring: Tay’s Meltdown proved once again what my university professor would say of computer science, “garbage in- garbage out”.

So on one side we would love to have computers help us with our CX, but on the other it looks really risky as any AI exposed to the public can be manipulated to reflect bad, racist or inappropriate responses to apparently innocent questions, sometimes just for the sake of it, others because of a specific sabotage schemed to bring it down.

Within Customer Experience, the relationship with automation has always been controversial. Would a customer like to be served by a robot and to what extent? Why would a company want to invest significant amounts of time and money to expose its front-line and most visible asset to malpractice and gruesome attacks from trolls and hackers?

The problem of any AI is, obviously, the learning. So probably the mistake from Microsoft was to trust the public network to be truthful and honest when teaching conversational skills to its bot.

Having worked in the customer experience realm for many years, I would never trust a bot to learn from public behaviour over social media: imagine your new hire agent sent to learn conversational skills and empathy…in the street?

but….on the other hand, I know that these guys (the CX teams) are literally sitting on a pile of interaction recordings that are rarely used, unless for some sparse quality management or compliance regulation. So why not use this big data to teach an AI, in a controlled, business-like though still real-life environment, how a conversation about your own brand or product should evolve? This idea might not be new but I haven’t seen anyone even testing it yet. Probably the biggest refrain is that AI projects are still in an experimental phase, are very expensive and bring little certainty of results.

But think about this: if you could have your new hire listen to thousands of hours of work conversations to learn how to address issues, how to talk to customers, how to properly escalate, how to behave in the interaction realm, and all in the business language of your own brand and company! This would be impossible for any human being, but for a bot…well, no big deal.

And the result: a perfectly trained agent ready to respond to your most difficult inquiries like your best skilled agent. Also, because every contact centre is different from the other, their recordings will result in different and more accurate learning and behaviour of the same AI. Isn’t AI in such case a dream come true?

As consumers, we probably would not care that the responses come from a bot, especially with digital channels where there is no voice and tracking a bot might be really tricky, and in the end, what matters most is the CX perception, not the reality. 🙂