Artificial intelligence (AI) promises to be the most transformational technology in half a century or more. Its adoption rate is faster than that witnessed in the early days of the internet as possible use cases continue to emerge.
To turn promise into reality, however, businesses need to act decisively.
Business leaders need to understand where AI should and should not be applied. They must appreciate the skills that are vital, both in-house and outsourced. And they need to understand the infrastructure and compute power required, too. That final point is particularly apposite when it’s considered that generative AI presents enterprises with the most demanding computational workload they have ever encountered.
It was with these imperatives in mind that Tech Monitor – in association with AMD – convened a series of roundtables across Europe to explore the potential of AI and what senior technology leaders should do next. The first event took place in mid-April at Balthazar, London.
AI starts with productivity
To kick off the evening, attendees were asked to identify what was top of mind when it comes to AI experimentation, adoption and management. A handful of themes emerged. These included identifying meaningful use cases, balancing the risk and reward of adoption, weighing up the benefits of first mover advantage versus being a fast follower, the relative merits of cloud and on-premise deployment, and the potential energy (and cost) implications of training large language models.
For most attendees, generative AI remains an experimental technology. The majority are “dabbling” with productivity tools as a proof of concept. Microsoft Copilot and similar productivity assistants are commonly deployed. Knowledge management, document management and sentiment analysis – especially in information-heavy service sectors – are other examples of AI use cases at the experimentation stage.
Why are most current applications of AI in this sphere? One attendee offered a theory. Ask a manager to consider AI’s utility and that manager will think about the personal need to cut down on time-consuming tasks. That’s fine as far as it goes, but if we are to uncover game-changing applications of AI, this attendee suggested, we require creative ideas from elsewhere. “We need to move one stage down [the organisational chart] to the operational staff, the doers.”
Uncertainty leads to opportunity
While most attendees remained optimistic about the potential of AI, a few were unconvinced. For example, an IT leader from a law firm noted: “My CEO will ask, ‘Why do you want me to invest one million dollars to solve a $10,000 problem?’ I don’t have an answer.”
If this group of attendees are representative of the wider economy then groundbreaking applications of AI remain elusive. That doesn’t not mean they are out of reach, however. Rather, more creativity is required – creativity that may provide competitive advantage. As one attendee working in financial services noted: “If my rivals aren’t sure what to do with it, then that’s an opportunity for me.”
Discussion of energy usage, efficiency and cost was threaded throughout the evening. Attendees expressed a desire to understand infrastructure energy use per-application. A ‘metred’ approach – accompanied by supplier-by-supplier benchmarking – would allow for better decision making and help reconcile AI adoption with ESG (environmental, social and governance) targets.
In-house or outsourced?
The conversation moved on to the interlinked conundrum of adoption: cloud or on-premise? Built or bought? On the latter, attendees were cautioned against relying solely on a commercially available Gen AI model. First, there’s a risk of corporate sabotage if cyber criminals access an organisation’s use of Gen AI. There is a future where bad actors choose model manipulation – and the subsequent chaos it might cause – rather than data theft as their most effective weapon. Second, those seeking to establish competitive advantage through AI are better off developing their own models and keeping their intellectual property under wraps.
One thing that is likely to influence the in-house versus outsourced debate is the availability of the necessary AI design and implementation talent. The UK is lacking those skills, an issue common elsewhere too. Education will solve the problem in the long term but doesn’t address immediate need. One attendee said finding AI architects was proving particularly difficult. Asked whether this would drive him into the arms of one of the hyperscalers (AWS, Microsoft Azure and Google Cloud Platform), he said: “No, it will make me look harder for the right people.” Others might take a different view.
So are IT leaders AI optimists or AI sceptics? Most appear to show signs of both characteristics. That mixture is likely to prove a suitable blend as they chart a path past the hype and towards AI use cases that work for them.
The next ‘Powering AI’s potential’ event, a Tech Monitor roundtable discussion in association with AMD, takes place in Copenhagen on 25 April.
