As organisations begin to turn AI-fuelled ideas into reality, what barriers remain in the way of implementation? How, for example, is existing infrastructure proving a challenge to delivery? What part will the cloud play in experimenting with new AI products? And if on-premise is likely to be part of the story, too, is the data centre itself in need of root-and-branch modernisation?
These questions formed the basis of a Tech Monitor executive roundtable convened in partnership with Lenovo and Intel in late June. ‘How AI and Hybrid Cloud are Reshaping Enterprise IT’ brought together senior IT decision makers from multiple industries to discuss their hopes for AI in the enterprise, and what work still needs to be done to ensure that their organisations can benefit. Here are the key takeaways from an evening of lively and open discussion.
Cloud vs on-premise
From the creation of sales and marketing assets to onboarding new hires, and from translation services to a knowledge base to serve internal users, attendees had little trouble thinking up use cases for AI in their businesses. Where these products should ideally sit within the enterprise, however, remained the subject of much debate.
For some, the answer is very obviously the cloud, given the geographic reach of the big providers in the space and the limited time-to-market required to bring AI solutions online. Others disagreed, arguing that cost efficiency and cybersecurity concerns meant that on-premise solutions were the best bet for supporting AI workloads.
Decision-making in the field tends to be more nuanced than these binaries might suggest. And sometimes, perceptions of a particular platform are set early on. As one attendee noted, marvelling at the trend to migrate workloads away from the cloud, the early migration to the cloud often ended in failure. Why? Rather than modernise applications to suit the new environment, organisations had simply carried on doing what they were used to doing on-premise. The “lift and shift” approach was always destined to failure.
Is the future hybrid?
According to one survey cited during the evening, two-thirds of organisations are looking at hybrid cloud solutions. Hybrid cloud combines public cloud (often in multiple varieties) with on-premise infrastructure, and is infused with the ease of use, elasticity, and scalability that has come to characterise the best of public cloud services.
Cost. Risk. Time.
When assessing the benefits of on-premise or cloud for AI workloads, IT leaders will make an assessment based on three key criteria, said a Lenovo voice around the table: cost, risk, and time. Often, two or more of those factors will play into decision-making. For example, one financial services firm that originally built a customer-facing service using the public cloud chose to move it to a private cloud environment on cost grounds. A subsequent change in industry compliance cemented the decision to move from public to private cloud, this time on risk grounds. Other attendees offered instances when public cloud delivered a more cost-effective solution to workloads where timeliness is an overriding necessity.
Why the hyperscalers will remain relevant
It was also suggested that geopolitical uncertainties – including those created by the second Trump administration – might force organisations to reconsider their dependence on US hyperscalers. Similarly, the introduction of the EU AI Act might force a pause in activity and motivate European countries, including the UK, still dependent in its trade of goods and services on the continental European market, to think about using technology solutions developed closer to home.
However, a straw poll of attendees failed to reflect this supposition. Just one financial services firm represented around the table said that political uncertainty had made them think twice about their cloud deployment. By contrast, well over half of the organisations represented at the event continue to rely on two or more of the three big US hyperscalers.
Concentration risk drives public cloud use
Rather than see a move away from the hyperscalers, regulation – and the Digital Operational Resilience Act (Dora), in particular – was seen by attendees as likely to drive greater use of public cloud. The act applies to financial services firms, as well as technology and service providers, and among its provisions is a mandate designed to reduce concentration risks. When it comes to public cloud, that means hedging their bets by deploying more, not fewer, hyperscalers.
The other dimension of the cloud versus on-premise debate is how either might apply to different types of organisations with different needs. Take, for example, small and medium-sized businesses that don’t have the capital to invest in hybrid solutions. Such firms, one attendee argued, are likely to lean heavily on the public cloud – a model that requires no upfront costs, affords the client the ability to scale up and down usage as required, and provides for a virtual team of security and infrastructure specialists. Similarly, those businesses that are born in the cloud are likely to need a very compelling reason to move some of their workloads on-premise.
Repatriation? Yes – except for (most) cloud-native organisations
For cloud-native organisations, with little or no legacy to worry about, the argument for sticking with their infrastructure choices is difficult to dispute. Repatriation to on-premise across the whole of the economy is becoming a growing trend, but it is much less prevalent among newer, cloud-first firms.
Famously, the virtual storage provider Dropbox moved from public cloud back to its own data centres in order to meet its specific application and scalability needs. For now, this reverse migration remains the exception that proves the rule among the cloud-native.
The workforce winners. And losers.
A final issue that occupied the thoughts of attendees is the impact AI implementation might have on future employability. How do organisations best manage the de-skilling and potential re-skilling of their workforce?
One attendee rejected the ‘man versus machine’ binary of this argument. Instead, they suggested that the future skills divide will be between those who know how to use AI and those who don’t. In effect, this was a rallying cry for prompt engineers – those who are able to interpret need and turn it into a request. “If you don’t know what you want,” said the attendee, “the AI can’t help.”
‘How AI and Hybrid Cloud are Reshaping Enterprise IT’ – a Tech Monitor executive roundtable in association with Lenovo and Intel – took place on Wednesday, 25 June at Balthazar, Covent Garden, London
