Even the biggest, most technologically-advanced businesses find artificial intelligence difficult.  There is a slew of examples of AI-gone-bad, where developers have accidentally built applications that have proved themselves to be racist, sexist, or seeming to advocate violence.

The problem is not with the technology itself, but with its human creators who so often bring their own unconscious biases to the table. Google, for example, listed 641 people working on “machine intelligence” – of whom only 10 percent were women.

This shows how even the biggest, most technologically-advanced businesses can make serious missteps in their journey towards better, more intuitive operations through AI and automation. How can businesses navigate this complex landscape and develop systems that are not only free from bias but which also develop real and measurable business value?

AI and automation – the human factor

Luminaries from Elon Musk to Professor Stephen Hawking have made dire warnings about the existential threat of AI to humanity. On a more prosaic level, many ordinary people have the erroneous (and dangerous) assumption that technologies such as AI, machine learning and automation will soon replace them in their jobs. In truth, we’re a long way from achieving this.

Current automated systems are still directed heavily by humans, with pre-set tasks and complex but defined and limited algorithms. There are no AI solutions in current business use whose actions are completely unpredictable, none that are capable of independent thought – and we see this as good business sense.

Rather than seeing the relationship between humans and these technologies as akin to that between master and servant, we should instead think of it more like a marriage. Automated and AI systems should be there to support us, not replace us. And, like a good marriage, they find firm foundations in dialogue.

The picture on the ground 

What does this look like in the real world? Well, to give one example, a business should not seek to replace call centre staff with full automation, but rather examine how they can use AI and machine learning to automate aspects of the process, making it more efficient yet still providing the human interactions that customers crave.

Before deploying AI and machine learning in contact centre environments, businesses need to decide precisely what they want to achieve. We would suggest that the most profitable aim would be to streamline the process, and work to shave a minute off each interaction. Companies can do this by categorising the different types of or reasons for customer calls, and then work out which low-value ones are ripe for automation – for example, with a virtual assistant.

But this should only be the beginning. Having automated aspects of customer contact, businesses should then be applying machine learning technologies and methodologies to calls to improve them still further. For example, they can identify why customers might drop out of their automated journey by analysing at what stage people left the call, and what patterns of interaction are most likely to lead to an unresolved query. They can then work to improve the process. This could involve manually building in more questions, or it could use machine learning to improve these models based on previous interactions (including person-to-person calls).

Cloud – the foundation for automation revolution

While the potential benefits of these technologies can, of course, spread far beyond the contact centre, this environment is a good illustration of the principles that will guide and underpin the automation and AI revolution.

Many of the new services that businesses develop will be built on natural language processing, such as Amazon Lex, Nuance, and Google Natural Language. These technologies can be deployed to “listen” to human-to-human customer service interactions, learning from human patterns of speech and then feeding into machine learning applications to make future calls smarter. Businesses can also use them to make the customer’s unique voice characterisation their password, as HMRC has done, to remove the need for tedious and hard-to-remember authentication processes.

The challenge with such sophisticated technologies is that their very complexity needs an extremely robust infrastructure platform. These aren’t just applications that need to be hosted: they have intense data processing, storage and security requirements, as well as integration with other corporate systems.

It makes little sense to host these complex and demanding systems on-premises; to work effectively, they need to be based in the cloud, where organisations can benefit from scalable or burstable compute and storage, along with first class security systems. The cloud also enables businesses to integrate AI, automation and machine learning technologies with other systems, applications and data into which they feed.

Revolutionary as these technologies will be to the customer experience (and much else), it shows how the cloud is actually the most transformational of all. As the foundation of tomorrow’s automation, AI and machine learning technologies, having the right cloud infrastructure to support your business’ ambitions has never been more important.