The Human Factor (in the Age of Machines)

All of this hype about machine learning must be addressed somehow. This blog post is about how marketers can coexist with machines, and not to leave full control to them. Too many human users are doing that already.

machine
“machine,” Creative Commons license. | Credit: Flickr by Jonas Tegnerud

No matter how far AI evolves in the future, for as long as humans remain as the dominant species on this planet, machines will exist to serve the benefit of human collectives, in some form or another. That is an optimistic view and possibly the best-case scenario.

Now, if we imagine the dark path as kindly illustrated in movies like “Terminator” or the “Matrix” series, AI may one day decide to eliminate humans as we are merely nuisances to them (the worst case scenario), or convert us into living, breathing battery packs to power them with our body heat (the next-worst-case scenario).

Even without such doomsday predictions, it is quite feasible that machines will take jobs away from most of us, starting with menial and repetitive ones and moving on to so-called white-collar positions with thinking involved. Not quite the end-of-the-world case, but definitely the end-of-the-world-as-we-know-it situation, as the cognitive process won’t remain as a uniquely human function.

Not too long ago, it was big news that AI decisively defeated one of the smartest human beings on Earth in the game of Go. It was quite an achievement — not necessarily for the machine, but for the humans who designed it. The machine, less than one year after that achievement, is now up to the level that its older version won’t able to match. The latest is that it doesn’t even play Go anymore, after having played the game by itself millions of times.

Here is my take on that event: First, why is that so surprising? Yes, the game of Go is far more complex than chess, with a virtually unlimited number of outcomes. But everything happens on a game board and the rules are quite simple. Machines and humans can observe and predict events within that set boundary. If machine does nothing but “1” task within the rule set for an unlimited amount of time without being bored or getting tired, of course it will beat humans who easily get distracted or grow tired.

So can we even call such a match fair? At some point in the distant past, a car passed the speed of the fastest human runner or even a man on a horse (with exactly 1 horse-power). But other than the fact that we still continue to humiliate horses by measuring the engine power in terms of “horsepower,” who cares about that? We don’t have runners compete against cars in the Olympic Games, do we?

The second point is that, yes, it is newsworthy that an AI beat one of the best Go players in the world. But so what? The history of computers has been a series of human defeats in terms of speed and accuracy since the very invention of the thinking machine. Computers have been outperforming humans in many ways all along, so why does everyone get so scared them all of a sudden? Is it fear of the unknown or loss of control?

We have learned how to coexist with clunky mainframes in the past, and we will learn how to live — and live well — with AI with or without cute faces. And that’s if, and only if, we maintain the “human factor” in the evolution of thinking machines.

So let’s stop thinking about how smart machines have become, and let’s think about what that word “smart” means.

What ‘Smart’ Means

Does it mean that it remembers things better than us? Undoubtedly. The best use of a computer is to have it remember what we don’t want to remember. Just because I can’t even remember my work number without my “smart” phone, that doesn’t mean that I became dumber. I will use the remaining memory space in my brain to store some other useless information, like the average driving distance of an old golfer or a name of an actor in some obscure movie. Then again, why even bother with all of that when I can just Google them anytime?

Barriers to Personalization

Recently, I stumbled onto survey results from marketers regarding “data-related headaches,” published by a reputable source. What surprised me the most was not the list of the pain points, but the way marketers expressed the severity of pains. They collectively answered that “moving data among different silos” and “gaining a

Recently, I stumbled onto survey results from marketers regarding “data-related headaches,” published by a reputable source. What surprised me the most was not the list of the pain points, but the way marketers expressed the severity of pains. They collectively answered that “moving data among different silos” and “gaining a single customer view” gave them the most severe headaches, while “personalization” brought not-so-severe pain. That gave me an “oh, really?” moment. Then they put “contextualization” (of data, I assume) and “turning data into insights, and insights into actionable segments” right in the middle.

From a data and analytics specialist’s point of view, it seems like marketers have no idea where the pains originated. Simply, proper personalization is not possible without the 360-degree view of a customer and insights derived from the data. So, in my opinion, the severity list seems completely backward. And it is just unbelievable that marketers “think” that they are performing some type of personalization without much of a headache.

During the past few months, I have been emphasizing the importance of personalization in modern marketing (refer to “Personalization Is About the Person”), and data and analytical steps to achieve such goals (refer to “Road to Personalization”). I’ve said similar lines many times before, but let me repeat: Proper personalization is not possible without understanding the target individuals as people. If marketers are thinking that buying some fancy software and putting transaction- and event-level data through it are the end of their jobs, they cannot be more wrong. Such activity often leads to “personally annoying people,” not impressing customers with relevant messages. If they were to automate such a rudimentary practice? Well, they are going to end up annoying their customers and prospects on a regular basis.

If you as a marketer are having a hard time stomaching what I am saying here, please then take a look at your inbox, which is filled with irrelevant messages — as it is for a consumer. Aren’t they filled with the kinds that you would purge mercilessly, as in “highlight all, then delete”? How many messages are really relevant and timely to you? Maybe one out of 300 to 400? Even the ones that are based on some tidbits of information that you left behind purposefully or accidently become really annoying after the third time you see the same darn message stemming from them. Ok, I get that some marketers think that they know me, but could they please not overdo it by turning on some expensive personalization engine on an autopilot mode from day one?

As I emphasized in my previous columns, personalization is about the person. Putting event-level or transaction data into a personalization engine is like putting unrefined oil into a high-performance engine. Not a recommended course of action, for sure. And don’t blame the engineer when things break down, even though the salesperson who sold you that engine probably claimed that it would make all of your marketing dreams come true.

Regardless, I think we can safely agree that personalization must start with the data. Unfortunately, not all data are created equal or are of the same quality (refer to “Chicken or the Egg? Data or Analytics?”). In fact, most data are utterly inadequate for high-level personalization that does not annoy people. So yes, the fact that marketers think that creating a single customer view out of all types of data from different silos is indeed important and difficult is a good sign. A critical change always starts with the recognition of a problem. It is just that marketers should never think that personalization engines could magically help them skip that painful step of data hygiene and consolidation.

If the data management were the first hurdle on the way to decent personalization, then the second challenge that marketers often face would be the analytical part of the journey. Deriving insights out of data and turning such insights into actionable segments require advanced-level analytical skills. Here again, automated machines do not perform the human part of the equation. Some marketers may have procured some automated modeling engine (again, with much fanfare as a magical tool). But who will set the goals of models and define the target for each model (refer to “Data Deep Dive: The Art of Targeting”)? Who will connect the dots between resultant personas and segments to actual offers and messages that customers and prospects get to see?

Even for cases where marketers must respond to a customer’s need immediately (e.g., for buyers who are specifically looking for a specific product right now), the rules of engagement (i.e., customer journey mapping) must be set up based on clear business objectives, as well as mathematical equations. Humans, not so surprisingly, can smell the sign of not-humanness from miles away, through even digital channels.