Artificial intelligence is set to transform businesses in every sector, but many CIOs are struggling to turn their AI ideas into practical applications. Ashraf Murtada, Deputy IT Director of Data Engineering and Delivery at HMRC, believes the key to success is replacing the mystique around the technology with practical explanations.
During CIO UK and Computerworld UK‘s Artificial Intelligence Summit at the May Fair Hotel in London on 28 February, Murtada drew on his 20 years of experience in software engineering and data analytics to give his advice on how to recognise the AI use cases that will benefit each individual organisation.
[There are 1,000 copies of the research available in an exclusive print run; registered CIO UK users can also download the full report here.]
Murtada, who leads a large scale engineering function at HMRC, delivering data engineering and capability, advanced analytics and focusing on cognitive technologies, said that the most pressing challenge concerning AI today is the aura of mysticism surrounding it. Organisations need to cut through the hype so they can understand how to apply it effectively.
Once any “magic” attributes are stripped away to express the essence of AI in simple terms, firms can start to identify practical business cases, explained a “self-confessed geek” with more than 20 years of experience in software engineering and data analytics.
Devising a strategy
A common definition of AI is a system that performs tasks usually reserved for human cognition and is able to predict outcomes and make decisions after recognising patterns.
When Murtada asked the audience how many of them had an AI strategy in place in their companies, only a handful responded positively, while a few more said that they were working towards developing one. However, the question that saw more hands raised was how many in the room were struggling with AI and didn’t know where to start.
“There’s a lot of promise with the technology, and that promise is making organisations very anxious and eager to implement the technology and try and deliver a huge business value that’s being promised,” explained Murtada.
He went on to expose how AI permeates every aspect from our lives in both the business and consumer worlds. Whether it’s in the recommendation engines used by Amazon or Netflix, Google’s Assistant, self-driving cars, robo-journalism and music composition, AI is now a part of our daily routines.
“There are lots of AI examples today; it’s far more common than we might think”, Murtada asserted. “There’s hardly any field today where you can’t stop to see real applications for AI.”
The economics of AI
Quoting Ajay Agrawal’s book Prediction Machines, Murtada said that “AI today is a lot like the internet in 1995.”
“People started talking about the internet as a new economy rather than a new technology,” the IT Director said. “But shortly afterwards, the impact all over the various aspects of life was of such an extent that technology starts fading into the background.”
With the expansion of the internet, the cost of digital distribution of goods dropped dramatically. What was previously done physically acquired a digital – and more inexpensive – form.
This is what in traditional economics is called downward sloping and demand curve, meaning a rational consumer will demand more of a commodity when its price falls: if something get cheaper, the demand for it will increase.
According to Murtada, that’s the case with AI today.
“When the cost of something drops, we use more of it. So the cost of prediction in terms of AI is dropping,” he said.
However, while the cost of something drops with an increase in demand, the cost of complements increase and that of the substitutes do down.
“You can see the economic relationship here: all of a sudden, all of your data assets become far more valuable, far more important, because they are inputs to prediction,” Murtada explained.
“Also training is better because this is experience, it’s factual knowledge, and the cost of complements to prediction – data and judgment – will go up. In addition, the cost of alternatives will go down: the cost of human prediction will become less valuable,” explained Murtada.
“This gives us an insight where we can shift our workforce so we can focus our workforce less on prediction, and more on these areas surrounding prediction. Because they carry more value,” he added.
The anatomy of a business task
During his presentation, the HMRC IT Director placed AI in a technology category of his own.
“AI is a truly special technology,” he said. “It’s considered by economists as a general purpose technology. If you go to any major tech event, you will find lots of technologies on display: IoT, cloud computing, mobile computing, blockchain – but AI is sitting in a category of its own.”
In Murtada’s view, AI is so special because the fall in cost enables applications in a wide spectrum of things that we do in business. In the case of AI, that thing is predicting.
“That is the essence of AI: it enables us to do very cheap predictions. Thanks to AI we can do them cheaper, faster and better,” he said.”If you replace AI with ‘cheap prediction’, it becomes less magical and more practical.”
With the cost of human predicting becoming less valuable, businesses should then discern where to shift their workforce, focusing less on prediction and more on areas surrounding it because they carry more value.
“If you go back to your offices and pick one workflow or process within the organisation, break it down into tasks, what activities are needed, and what decisions are needed in each task,” Murtada advised.
“Then you will see that you can start in piecemeal fashion in implementing automation for predictions that feed into those workflows attention, so you don’t have to make a complete overhaul of the business structures to make it smarter, you can just target the decision points with prediction.”