Thinking for business: how AI has a role to play in everyday commerce

Artificial intelligence technology is behind seismic breakthroughs in science and medicine, but it is also available to all kinds of businesses

Aidan Connolly, chief executive, Idiro: ‘Since Covid we've seen a major uptake in customer-oriented use of AI’

These days, artificial intelligence (AI) and machine learning (ML) is a hot topic with businesses asking how it can add value to their operations, but the uptick in interest in recent years obscures the fact that, in fact, it has been at work for a long time.

The first industry to really benefit from the practical application of AI and ML technology was telecommunications, said Aidan Connolly, chief executive of Idiro.

“The telecoms sector, I think, was among the first movers, along with banking, albeit for slightly different reasons. Telecoms had a lot of data. Eircell, Vodafone and O2 were all using AI and ML in the 1990s and early 2000s, and banks were looking at it in terms of risk profiling of their customer, things like to whom they should give loans,” he said.

Today, things are very different, and very obviously different, with AI having a clearly visible impact on our lives. However, the fact that we can now see AI at work – in voice assistants, in chatbots, in recommendation engines, and even in partly-automated transport – does not mean it is only now becoming important.

“It came to public notice when companies like Amazon and Netflix’s recommendation engines were put to work recommending movies to people, or saying people who bought this also bought that. Before that, you would have felt the impact but not necessarily known,” said Connolly.

There is more to AI than recommending products, however, and today even smaller businesses can deploy it.

This is because, firstly, compute, connectivity and storage have all collapsed in prices and, secondly, it no longer requires massive capital investment and does not require specialised hardware and software.

“The small- and mid-sized companies now have an opportunity, as do more conventional businesses like food manufacturers who can use it to optimise processes,” Connolly said.

Typical applications for small companies include using it to understand their customers and optimise customer experience.

“Certainly, since Covid we've seen a major uptake in this kind of use,” said Connolly.

Other tasks include scraping social media and customer communications to get a sense of how people feel about the business.

“You can also do things like discovering positive or negative sentiment quite easily, but to improve it is the challenge,” Connolly said.

At the cutting-edge, AI is being used in life and death applications.

“What we’re seeing is things like drug discovery, and the ability to simulate biological organisms or analyse medical imagery. Seeing how cells and biological entities interact with drugs is very exciting as it will rapidly reduce drug development times,” Connolly said.

However, there are issues that need to be considered. For a start, AI models are only as good as the data that goes in.

“People focus a lot on the models but not enough on the data, and this could be a real problem. AI is used to shortlist candidates for jobs, which can perpetuate biases against people who are members of under-represented groups. You could have people being discriminated against in insurance premiums or in loan decisions,” said Connolly.

Idiro has set up an AI ethics centre and is now developing a product to make sure AI is ethical. This is not a niche concern, either. The EU is introducing the AI Act, which will be like GDPR for AI, meaning companies will have to be compliant – and have their AI audited by third parties.

“If you call a bank or an insurance company, they have to be able to explain the reason for a decision. Saying ‘The model told us no’ will not be enough,” Connolly said.

In the meantime, companies should not be intimidated by AI, he said. Instead, they can get started on deploying it with small, coherent projects.

“Chances are, at least one of your competitors has jumped in. I would encourage companies to take a measured approach: don't just hire a data scientist; try and find out what you want to do and then identify some quick wins or low-hanging fruit. If you do it gradually, that will give you confidence to continue, and it also allows you to manage your budget,” he said.

Working in this way also will also allow for the AI to prove itself, including to internal critics – and there will always be naysayers, according to Connolly. Indeed, AI can be a tough sell in some boardrooms.

“Many people in management get to the top on the basis of their gut, their acumen, their connections, but not on data, so moving to data can go against everything they know,” he said.

Against that, there is the fact that digital transformation has left many companies floundering. A well-planned and well-executed AI project could turn this around, driving real value for businesses.

“Most digital transformation projects are incomplete. Yes, they transform digitally, but they have no plan, so they end up with all this data online but it may still be siloed. Instead of that, you can have a data strategy that doesn't cost the earth,” Connolly said.