"Ground control to Artificial Intelligence: come back to Earth"

artificial intelligence space.jpg

Tim Coulthard looks at the need for Artificial Intelligence to find a more pragmatic role in mainstream industry as its early claims make way for a more accessible form of technology

Artificial intelligence has had several dawns over the past decades as the technology has evolved from the stuff of science fiction to the more implementable, accessible science seen today.

But in its current era, there have been lofty promises, from revolutionizing cancer treatment to transforming transport though autonomous vehicles. Those aims may well come to fruition in the coming years, but while they remain unresolved, they hang over the technology, used by some critics to brand AI as over-hyped and over-ambitious.

Looking around the AI landscape now though, the industry use cases are more focused and smaller in scale - what we might call ‘pragmatic AI’. I spoke recently with an enterprise AI provider, who conceded that early claims had pushed AI out of reach in the minds of potential customers. Their new business development strategy is now based on smaller, narrowly defined tasks like calibration and quality control that generate rapid, quantifiable ROI.

The space exploration title of this blog is inspired in part by our recent podcast with Professor Tom Davenport, where we discussed in depth the concept of ‘moonshots’ versus ‘low-hanging fruit’ in the context of AI. In his new book, The AI Advantage, he cites the example of NASA, an organization previously capable of literal moonshots, but currently using AI largely for lower-key tasks like HR administration and accounting automation.

While the great leaps for science will surely come, the AI scene has lowered its sights, at least publicly, to help it gain traction and credibility. Take chatbots: this entry-level technology can be implemented rapidly at minimal costs, but puts organizations in a place where they can legitimately claim to be ‘doing AI’. The barriers to entry are down. The technology now allows companies to automate the ‘busywork’ of basic tasks that takes up increasing amounts time for skilled employees.

Of course, current potential extends beyond the ‘toys’ of chatbots, to serious industrial applications. GE Digital has found success with its Digital Twin offering, where AI and analytics are pushing boundaries in predictive maintenance. Digital models, ranging from replicas of single components through to entire production facilities, map performance and learn to predict maintenance requirements in advance, reducing downtime and costs from unexpected stoppages.

The first projects using AI may not ultimately be transformative for your organization, but getting started breaks down the first cultural, psychological and operational barriers

Our recent articles with Johnson & Johnson have explored developments in demand sensing where AI algorithms process huge amounts of data to create new levels of predictive demand, revolutionizing the potential of manufacturing and supply chains to deliver right-place, right-time strategies. The mantra of ‘store data, you know never when you’ll need it’ has echoed around industry for a while, and that philosophy is finally making commercial sense, as organizations combine their sales, inventory and production data with external geospatial data. Even Amazon, a front-runner in AI, has largely followed a pragmatic strategy of what Jeff Bezos call ‘quiet but meaningful’ uses.

What’s clear is that the time to start is now. The first projects using AI may not ultimately be transformative for your organization, but getting started breaks down the first cultural, psychological and operational barriers. Those moonshots may still be a way off, but knowing how a rocket works is a vital first step.