Is AI Failing Businesses?

The power of AI is potentially awesome - but is it fulfilling that potential yet?
31 August 2022

AI could power the next business revolution – but not yet.

AI (artificial intelligence) makes great science fiction. Always has done, since the dawn of the genre, long before anything even vaguely approaching AI existed in our real world. But since that’s the world in which we live, do business, and look for distractions to quell the nagging ache of disappointment deep down in our sci-fi souls, it’s worth looking at an inconvenient question.

Let’s ask another: will AI continue to fail businesses for the next 5 years? 10 years? The indefinite future?

To both questions, there’s a simple answer, which is to take a deep breath, say yes, get on with our day and binge that new Netflix show about the artificial intelligence that plans to take over the world.

It’s not, technically, that we’d rather live in the Netflix world where an artificial intelligence could take over the world. It’s just that we’d like artificial intelligence to live up to the lifestyle – and business – promises science fiction has always made to us for such an intensely revolutionary technological breakthrough. We’d quite like to live in the world where it’s plausible that AI could destroy the world, rather than the world in which Alexa clams up if you ask it to play a song it doesn’t recognize.

Low-Drama AI

One thing that leads to AI failing businesses is that we have very muddled and limited expectations of what AI could do for businesses. Our science-fiction imaginations always put artificial intelligence into dramatic circumstances – it’s always in self-driving robo-taxis (that can probably fly), or realistic androids that wait till the second act to go from just slightly creepy to murderous, or, legendarily, it’s in the life-support system of space stations we’ve somehow built and made autonomous of a ground crew back on Earth.

There has not been a corresponding literature (with perhaps the exception of Philip K Dick’s Autofac) that shows AI in a believable business setting. So, developers determined to use artificial intelligence in a business setting have to invent ways in which it can revolutionize the work of the business from scratch – which is hard to do when faced with a human-based system that has worked as well as it knows how to do since time immemorial.

AI has yet to find its killer application in business. Certainly, it is now used to deal with a large percentage of customer inquiries – the chatbot has become particularly ubiquitous, making an impact on the customer service realm as a whole, and new AI assistants have evolved the technology. But it has yet, for instance, to close the gap on humans when it comes to live escalation queries. In those scenarios, it still has to “transfer you to an adviser.” That’s an example of AI letting down business. In fact, in technical terms, it’s an example of “lo-fi AI + human” functionality. That’s a combination that may well be the best we can expect from AI for at least decades to come. It delivers on the restricted premise of its promise, but can’t close the gap and so needs human intervention to go “the last mile.”

The Last AI Mile

The last mile is what we want from AI. At the point when an AI can do that, it will at least fulfil our expectations. Robo-taxis would be hi-fi – entirely automated, so the human input is as close to zero as it can possibly be. What we’ve discovered in striving to get to automated cars is that lo-fi AI + human can give us an acceptable level of advancement, while striving for hi-fi AI leaves us facing a currently insurmountable challenge, because the level of precision you need to have an AI control a car in terms of split-second decision-making is beyond our current level of AI development. And probably will be for decades. That’s where the disappointment kicks in: the exciting, the killer, the sci-fi apps which would make AI really rock our world all exist in hi-fi – and hi-fi AI is infuriatingly beyond our grasp right now, even though we’re close enough to practically taste it.

As well as having to imagine from scratch the ways in which AI could revolutionize the business world, there is that same disconnect between the possible, lo-fi AI + human, and the impressive, hi-fi AI, able to complete the whole of a task without human intervention. At which point, as with previous revolutions, the question arises – what do we need humans in the system for at all?

The Business Siri

There are platforms being developed that should at least AI a fundamental function in business, like Xena, an attempt to give everyone in any given company an AI personal assistant. But Xembly, whose idea Xena is, is keen to say it won’t be arriving any time in the particularly near future. And even if it did, the idea of it becoming a business-revolutionizing killer app for AI is unbelievable, because bottom line, the idea of typing commands to an artificial intelligence, that then sends your form emails, when you could have sent your own in almost the same time and maintained at least the illusion of human connection, is not going to set the business world alight. And also, the fact that you still have to instruct the AI means it doesn’t get above lo-fi AI + human.

So, AI is currently failing businesses – in a way which, for instance, robotics didn’t – partly because re-imagining the world of business augmented by artificial intelligence is harder than it was with industrialization or robotics. With industrialization, the aim was to create machines that did the physical work of humans, but faster and more accurately. Invent a better loom and you revolutionize business, while putting hundreds of thousands of people out of work. With robotics, you took the skilled elements of a job, wrote a program, developed a robotic working machine that could follow the program, and revolutionized business… while putting hundreds of thousands of people out of work.

As a result of both previous revolutions, the skill of a human in a system – the reason to pay that human a wage – has become more and more a matter of their intelligence, their knowledge, and their judgment. While lo-fi AI + human remains the best we can achieve – as it looks set to do for several decades – the development of the killer AI app for businesses looks unlikely, and we’ll have to put significantly fewer people out of a job.

The Universal Translator?

That said, there are stirrings in business circles about products that might just qualify as a low-key killer app for AI. Things like the new Google Glass, which can offer real-time or near-real-time translation of visible and audible speech from the language spoken to the language heard or seen. That has the feeling of properly sci-fi hi-fi AI (from Star Trek’s universal translator to The Hitch-Hiker’s Guide to the Galaxy’s Babel Fish). That would open up international markets without having to learn new languages or hire interpreters.

But for the most part, artificial intelligence is a way off yet from being a trusted technology at the level it would need to be to fully satisfy businesses and help them on a day-to-day basis.