In my last post, I made the case that organizations need to start automating data management. In short, it's not 1985 anymore. It's silly to think that we can continue to manage new, complex data sources and types with manual processes and tools rooted in the nascent days of enterprise computing.

Today I'm flip-flopping. I'm going to explain why embracing full automation is equally silly. Before I do, though, it's critical to recognize the perils of AI.

Consider a recent piece in The New York Times about YouTube serving up increasingly risqué videos of teenage girls to pedophiles. At first glance, you might wonder: How could that possibly happen? Did a rogue employee write code or tweak the algorithm to do this?

No – no individual programmed YouTube to make these repugnant recommendations, But here's the rub: no one programmed it not to either. The algorithm just did its thing: served up what it thought were videos that people were likely to watch. More videos means more ads watched. This means more money. That is why YouTube is arguably worth $100 billion.

And that, in a nutshell, illustrates perhaps the scariest downside of automation.

Think about it. You don't have to be Ray Kurzweil to realize that AI will become increasingly sophisticated during our lives. Still, we're a long ways away from even approaching the g factor, never mind superintelligence. For at least a little while longer, we humans will be able to do many things that machines simply can't.

One of those things is using judgment to solve data management issues.

An example

For instance, consider the fictitious ABC Distributors – a large restaurant supply company with thousands of customers (read: diners). ABC's management buys into the upsides of fully automated data management. (Hey, maybe its faux CIO read my last post.) As such, it configures its internal systems to eliminate any human involvement. ABC brands its new creation autoDM. Without the need for people to manage its data, the company furloughs the folks previously charged with data management. Another example of successful automation, right?

But there's a problem. Someone forgot to tell autoDM that ABC counts as customers 14 different Main Street Diners. (There are more than 50 on yelp.com alone.) autoDM consolidates the different master records without giving it another thought because it doesn't really think.

To use a technical phrase, all hell then breaks loose. The Main Street Diner in Plainville, CT receives 14 times as many napkins, replacement forks and spoons and cleaning supplies as it normally does. Its owner doesn't know where to even store the stuff.

Meanwhile, the other 13 Main Street Diners never receive their supplies. Bathrooms are filthy because the paper towels never arrived. Customers complain and vent on social media. Some write negative Yelp reviews. Their managers have to run to Costco or Walmart in the middle of busy shifts. ABC senior management scurries but it doesn't know how to write SQL statements and fix the problem.

The end result: All 14 Main Street Diners take their business to another restaurant supply company – one that hasn't automated its data management.

And that's just one potential drawback of fully automating a business process. Perhaps a more common consider is how employees will react. There's a reason that the term Luddite remains popular. (For more on this, see Automating Workplace Tasks Can Backfire if Employees Shun the Technology.)

Simon Says: Find a middle ground.

Sure, new automation capabilities can improvement organizations' data management efforts. I'd stop way short, though, of removing people from the equation altogether.

Instead, try a hybrid model similar to one described in a recent Washington Post piece on WalMart:

...the technology can only do so much. When the AI senses a problem, it sends an alert to the handheld devices most Walmart workers are expected to carry, saying it is time to corral the carts or replenish the produce. The store’s roughly 100 human associates are the ones who do the physical work.

Feedback

What say you?