Semantic technologies of various flavours have the potential to discover connections and enable insights that are powerful, valuable, intriguing, insightful, and surprising. However, as with so many technologies, there’s a flip side. As tools grow more capable and data sets continue to blossom, it becomes ever more likely that the segmented lives of web users — whether public and private, law-abiding and nefarious, or respectable and risqué — will be joined up without our explicit consent. As developers who see the best in people build ever-richer tools, technology companies that cannot — or will not — understand the value of the ageing legal system’s checks and balances continue to use them in pushing the boundaries of what society considers to be acceptable. There are already scare stories, and as the reality becomes more capable, those scare stories will undoubtedly become increasingly terrifying. Some of them may even turn out to be true. But it’s not necessary to believe the full Orwellian horror of a Mountain View search engine (or a Menlo Park social network) that knows everything about you, and acts upon that knowledge. Even in the smaller every-day injections of semantic smarts into business processes, there are things that should perhaps make us sit up and take note. Just because semantic technologies can discover patterns in customer behaviour, doesn’t mean we should necessarily act upon them too soon.
Bill Guinn of Amdocs delivered one of the keynotes at the Semantic Technology Conference in San Francisco back in June. He painted a picture of a $3 billion company in the business of delivering customer experience systems, managing customer accounts and providing the first line of support for clients in call centres around the world. He talked enthusiastically about the way in which semantic technologies “improve revenue, reduce cost, and improve customer satisfaction,” and went on to assert that “I truly believe that semantics can be a game changer in just about any B2C model.” Guinn showed example after example in which existing processes became smarter and more cost-effective. He persuasively demonstrated the value of semantics to his business, and to his customers. From what we saw, it appeared to be a text-book example of getting these things right.
But Guinn described one small use case that made me uncomfortable. It wasn’t wrong, necessarily. It certainly wasn’t bad, or malicious, or evil, or any of the other emotive words we use when technology goes too far. But it seemed unnecessary, and it felt — to me — like a step too far. If I were the customer, I think I’d be profoundly unsettled if it happened to me.
So what was this terrible crime? It was, quite simply, a proactive intervention by a call centre agent, enabled by semantic technologies. You can hear Guinn describe it in his own words during the 5 minutes after 7 minutes 21 seconds into this video.
Using prior knowledge and context to route my call to the customer service agent I’m most likely to want? Good, useful and proper. Using prior knowledge and context to ensure that the call centre agent has details from an ongoing case file up on their screen before I need to ask for it? Useful, and efficient. Using prior knowledge and context to juggle the order of options presented by the nasty “Press 1 for…” machine? Probably pretty useful, unless I call so often that I’m reaching for “2” before you’ve told me what it does this week. But using prior knowledge and context to greet me with “Hello Paul, I know what you’re calling for” ? Downright creepy. And it smacks of smugness and hubris on the part of the call centre agent. All sorts of clever back-end processes are at work to mine my customer history, to flag any oddities in my account, to identify potential upsell opportunities, and to ensure that all of this is available to the call centre agent in a timely and digestible fashion. That’s all great, and will probably give me a better customer support experience and make Amdocs more money. But the information should be there in the background, waiting to be called upon after I’ve explained why I’m calling. “Ah yes, I have that information here” is a much better response to me calling and asking about a surprisingly high bill than Guinn’s suggestion of a preemptive “Hello. Are you calling about your bill?” Don’t be creepy. Let me feel like I control the call. But then go on to extract all the value that Guinn describes additional context as providing. It’s not a technological problem. It’s a human one.
Sometimes the people with the clever technologies, and the people with the stopwatches and calculators that strive to make every little interaction more efficient? Sometimes, those people need to take a step back from the technologically clever, the efficient, the cost-saving, and think about how it would make that little old lady in Des Moines or Kirkcaldy actually feel.
And I think she’d feel spooked. I certainly would.
Just because we can, doesn’t mean we always should.
Semantic technologies offer a powerful set of tools, and Guinn described myriad ways in which Amdocs are putting those tools to use in creating a more efficient, a more profitable, a more customer-satisfying business. It sounds great, but the creepy mind-reading thing on the phone has to go.
Image by Flickr user ‘CBS Fan’