The Best Way to See AI: Stop Looking

Today we’re going to put AI in a different genre: the scary-monster-isn’t-so-scary genre, in which the “monster” turns out to be an ally (like in Terminator 2: Judgment Day without the tearful goodbye into a vat of molten steel), or a misunderstood good guy (like in Beauty and the Beast

So let’s take a look at the current AI craze, its effect on organizations today, an approach to consider instead, and a practical way to get started.

The message AI is sending

Back to the AI Revolution for a moment. Like any self-respecting uprising, AI seems to be calling for a full-throttle marshaling of the troops — a multi-theater effort to devise a plan, gather supplies, rally the recruits, and proceed to conquering.

What this actually looks and sounds like for organizations right now is a worldwide alert across all markets and industries to develop a comprehensive AI strategy, collect a flood of data plus the tools to tame it, amass data scientists and other analytical talent, and compete or die on a sleek new playing field where your hardscrabble infrastructure no longer fits.

Every day, this alert blares. AI experts, analysts, and pundits proclaim the questions you must ask before embarking on an AI initiative, reveal the signs that you’re ready for it (or not), list the complex steps you must take to achieve AI maturity in [name a year], advise you on the career moves you must make in light of its onslaught, and laud the companies doing it well (read: better than you are).

It’s all well-founded advice. But how do you take on an apparent monster like this?

The “frenzy or freeze” reaction AI is creating

Turns out a lot of organizations are not taking it on. Facing a phenomenon that has been called — by true experts, without irony — “another Cambrian Explosion” (the period 540 million years ago when higher forms of life suddenly appeared) and “the new electricity” (not as a metaphor, but as a reference to the literal invention that transformed the world in 1879), organizations commonly react with a “weight of the world” attitude: they “believe [they] have only two choices — a) world-shifting innovation, or b) do nothing.”

This attitude plays out in the data. A current, exhaustive study by the Boston Consulting Group and MIT Sloan Management Review found that while 84% of businesses in 112 countries and 21 industries believe that AI will give them a competitive advantage, only 23% are using it today.

Here’s the thing: the arrival of AI really is a colossal event of mind-bending proportions, and equating it with our planet’s most consequential turning points is not hyperbole. But obscured somewhere in the thick forest of research, literature, and marketing around AI is the fact that it is not an ominous, advancing army that demands a sweeping overthrow of everything you’ve built.

It can be more like a small tactical unit that swiftly and nimbly liberates you from your oppressors of heavily manual tasks and their accompanying bottom-line costs.

Military (and monster) metaphors aside, AI can be approachable. You can explore it in manageable chunks. You don’t have to rip out your whole technological or organizational foundation to get started, and you don’t have to avoid getting started (which you know isn’t an option much longer, anyway).

An alternative reaction: hop on and enjoy the ride

Taking AI for a spin in a specific department, or for a specific task — as opposed to transforming your entire organization into an AI-enabled racetrack first — is doable, and even preferable. It’s in line with a “micro-innovation” approach that uses incremental, pragmatic gains to build a digital ecosystem of competitive advantage, with little risk.

It’s also in line with the nature of AI, a technology that can lift burdens and relieve bottlenecks without also introducing them.

In fact, making things easier, more useful, and even more fun is what AI is supposed to do. And it can do this by just showing up — amiably, without your bracing for impact — to fill the gaps you didn’t know were there.

We’ve seen this on the consumer side with products like the now-iconic recommendation engines from Amazon and Netflix, natural-language-powered digital assistants, and apps that enable you to organize photos with no identifying labels. Behind all these marvels and others like them is, of course, the subset of AI known as machine learning, and the subset of that called deep learning — algorithms that teach themselves by learning the patterns in vast quantities of data.

Individual consumers don’t need to prepare for these algorithms with feats of technological or organizational heroics. To begin reaping benefits, consumers need only to engage with the product built on top of the algorithms. Or you could say that consumers need only accept a turn on the ride. And this approach shouldn’t be limited to individual consumers.

A new way to view AI: stop focusing on it

How can organizations simply engage with AI, like consumers can? Because AI itself isn’t the end game, and shouldn’t be the focus for any organization. AI is merely a vehicle for value.

Jeff Bezos, who knows a few things about AI, acknowledges that it’s a renaissance. But he also calls it something a lot less dramatic: a “horizontal enabling layer.”

“A lot of the value that we’re getting from machine learning is actually happening beneath the surface,” Bezos says.

In this way, AI is like design: when done well, it doesn’t announce itself. Rather, it directs your attention to the product it’s supporting. Put recently by Ruthia He, a product designer at Facebook, “Good design is invisible. Why? Because to design is to find a way to solve a problem.”

Likewise, good AI is a seamlessly-adopted solution, or readily-enjoyed feature, with an unobtrusive AI foundation.

“Good design should also be self-explanatory,” noted He, “and make people feel like this is just how things should naturally work and there is no better way to do it.”

Good AI — useful AI that provides the competitive advantage it promises— should be the same: a product that performs like you expect it to, and delivers results without friction or fluff.

The fact that good AI is also unassuming AI is why some experts — even the ones who equate AI with with the dawn of life on earth — also say that in a few years, AI won’t be anything to trumpet. Instead, it will be expected. And then, brands won’t be able to lead with “powered by AI” because everything will be powered by AI. So brands will have to lead with the real value that their products deliver.

So how can organizations get started with AI?

Organizations today have options for entering the AI arena and becoming immediately productive without gladiator-style prep or algorithm-wrestling. Many products take on well-known and universally-experienced business problems, and solve them with AI, machine learning, and deep learning.

These kinds of products often put AI at the core of an automation solution that elegantly removes all the futuristic fanfare (or doomsaying) along with a specific set of pain points, inefficiencies, and redundancies.

And this kind of automation does not obliterate a hard-won infrastructure or displace workers. Instead it works with and unburdens stressed systems, and serves talented employees by alleviating routine and repetitive work — physical, yes, and cognitive too, like legal discovery and financial reporting.

For example, at Nousot, we looked at core business analytics — things like forecasting, clustering, propensity, and optimization modeling — and realized how well-suited they are for automation. They’re very manually intensive: organizations routinely employ dozens of data scientists to build and refresh models, a process that takes weeks or months, with results that soon become outdated and lose their value.

This need for hands-on talent is exactly what keeps data scientists from getting to the more creative applications of their jobs, and smaller businesses from maturing in their analytics capability.

Another thing about forecasting, clustering, propensity, and optimization modeling: they require highly advanced math functions, but they also require the same ones repeatedly, like variable qualification and selection, model technique selection, and model development and testing.

Doing advanced math over and over again? Artificial neural networks have been sitting around since the 1950s just waiting for the compute power and the data to own that job like a boss.

Giving data scientists their time back, and giving teams without data scientists a play at sophisticated analytics, are examples of AI at work as it should be, and can be: invisible, self-explanatory, and solving everyday business problems so cleanly that you can’t imagine things any other way.

Does this mean that you can gloss over a comprehensive AI strategy? Probably not. But it does mean that you don’t have to overcome any monsters in doing so. You can deploy some automation and start tackling the real monsters, like the growing costs of a protracted analytics process, or of skilled data scientists not using those skills, or of anemic market insight.

You can stop looking at AI, and start seeing real benefits by focusing on the solutions it enables instead.