Where does AI go next in businesses like yours?

One of the most notable learning lessons about the AI hype-curve is that this is not a one-trick pony.

Most solutions that businesses are curious about could be characterized as pedestrian. This is where AI has potential to be applied to solve business challenges, and grow the potential of incumbent business models. Nevertheless, there are those possibilities with the accelerating potential of AI that bring with them risk of existential threats that not even politicians can ignore.

As I write this post today, the great and the good of the AI industry, and representatives of governments from around the globe, are camped out in Bletchley Park (the home of the Colossus computer and the WW2 code breakers) to attempt to frame what ‘safe’ and ‘not-so-safe’ AI looks like.

While—yes—the Orwellian Terminator and I-Robot future lies down one fork in the road, the other fork holds countless opportunities to safeguard and improve human life, prosperity, parity for all, and maybe even the promise of saving the planet.

To be clear, the vast majority of AI projects being discussed in the business world today fall safely into the pedestrian camp. Governments, particularly the UK government, is working hard to circle to wagons and formalize the rules of the AI game to safeguard society from the negative aspects of AI applications.

Let’s hope they do. In the meantime, the evolution of AI without boundaries of regulation steams on, risking the creation what some fear will be the AI equivalent of the ever growing layer of satellite debris that now skirts the earth around 2,000 km above the surface.

Of pedestrian AI, we are seeing three groupings of AI evolution. These are well articulated in what Roy Bahat, Head of Bloomberg Beta, so eloquently describes as his Loom, Slide-rule and Crane model.

‘Loom AI’ describes application use cases where AI is being used to drive transformation of manual tasks at scale that no human really wants to do, even when paid a salary to do it.

‘Slide-rule AI’ describes use cases where AI empowers humans to leverage AI to get their jobs done better, perhaps equipping human workers to perform tasks that were previously dangerous to do, or improving accuracy, scaling out decisioning, generating content, etc. This is most typically where generative AI sits.

‘Crane AI’ describes use cases where AI does things that humans could not previously have done. This is often where deep learning models sit.

It’s easy to get caught up in the AI hype bubble. The surest way to avoid doing so is to follow this three-step plan and build it into your 2024 budget.

Start cavorting around the tech market in search of useful AI solutions and you risk your team getting lost down rabbit holes ‘finding work’ for AI. I would say a better approach is to understand and prioritize challenges and problems that AI has the opportunity to solve.

Acquiring this insight is best done by engaging with your workforce, sales channel, customers and supply chain to explore opportunities. For example, what manual tasks are key roles performing that ‘take the time?’

With so many potential roles for AI across your enterprise—and with the added potential of Grant funding and investment from outside your enterprise—business leaders would be remiss to ignore it’s potential, even if immediate returns look hard to put a number on.

It makes sense for ‘someone’ to be assigned overall responsibility over your AI project to avoid the inevitable likelihood of every department or team kicking off its own pet project. Do it now, before AI adoption builds a head of stream, and you won’t be picking up the ‘after the party’ pieces of a fragmented and potentially unregulated AI free-for-all.

In the mean time, the most likely barriers to full-scale adoption of AI across your business are expected to be these two factors:

Data is fragmented and not composable

Most IT projects (even AI ones) fail due to data quality, data accessibility, or issues of fragmentation resulting from data being held within source applications.

Before deep learning can do its thing, AI systems must have access to data from across your business data landscapes.

Talent shortages

Nobody can agree on how many jobs will come and go as the result of AI. What we do know is that roles engaged in manual data processing, low level decision making, customer service support, or low level content origination are going to disappear fast, and at scale.

To temper this bad news story, we must also recognize that essentially every knowledge worker role will be supplemented by AI in the next decade. And, indeed, many job roles will be created by the advance of AI that we can’t even comprehend right now.

Stepping back to Bahat’s Loom, Slide-rule and Crane model, we know that talent will not only be restricted to core programming and development of AI systems. Harnessing ideation with AI will be equally important and require equally hard to resource skill sets. We can also expect to see an ‘us and them’ divide between younger generations embracing AI, and older ones wishing to bury their heads in the sand.

Sure, AI has made a splash—but so it should! It is after all impressively game changing and no-one can ignore it. It is the .COM bubble of the Millennial generation.

Where does it go next in businesses like yours? Like all big ideas, humans have a habit of looking for clever ideas. when sometimes the obvious answers are more boring (for sure) and more effective.

To my mind, the most impactful AI solutions in terms of business returns will come from generative AI and micro-automations that AI can solve relatively rapidly in every functional department. It’s about kicking the mundane and tawdry work into touch. These micro-automations hold great promise in marketing, sales, finance, HR, information security, compliance and many other areas.

Opportunities for deep learning and game-changing ‘AI-Crane’ innovations are sure to happen, but data (as always) tends to be the sticking point. Before you can truly harness deep learning, questions need to be answered on how to make data composable, work out who owns it, and whether information security and data privacy implications can be resolved.

In the short-term, it would be a safe play to focus on the quick-wins that come from micro-automation, encourage open mindedness and ideation with AI, and prepare your enterprise for the storm that’s sure to come.