AI adoption drives cloud rethink for organisations

As AI moves out of experimentation, Macquarie Cloud Services is finding that data risk and cost are key considerations.

AI adoption is shifting from experimentation to practical business decisions, with customers increasingly focused on the realities of deploying and managing AI.

Two concerns have come to dominate customer conversations: data risk and cost, according to Phillip Wallace, business manager, Macquarie Cloud Services.

Organisations want to know where their data resides, whether it is sovereign and protected, and whether deployments meet compliance requirements, said Wallace, speaking at this week’s Dell Technologies Forum in Sydney.

At the same time, they’re struggling to forecast AI expenditure as token consumption and model costs rise. In some cases, early AI deployments often defaulted to frontier models, causing costs to escalate, and creating huge uncertainty around projected costs.

"CFOs are asking the question now for the first time. But IT people don't know how to forecast what their spend is going to be. They're churning through tokens at a rate they never could have predicted,” said Wallace, at this week’s Dell Technologies Forum in Sydney.

The solution is to consider model choice, infrastructure and workload placement, rather than automatically running everything in the same environment. But there is no one-size-fits-all infrastructure model.

“End-user experience matters, customer experience really matters. Risk is a big thing and cost is a big thing. You’ve got to get the balance between all four of those pillars correct, and that will govern what the infrastructure looks like,” he said.

Macquarie specialises in servicing mid-market organisations and Wallace said they face particular challenges. Unlike large enterprises, they often lack the teams, resources and investment capacity to manage every aspect of AI themselves.

“Many mid-market organisations are waiting for their SaaS providers to introduce AI capabilities rather than developing their own strategies,” he said.

While this can make AI adoption simpler, it also potentially locks organisations more tightly into a particular platform, he explained.

Wallace recommends organisations start with the desired business outcome, not the technology. He also advises against attempting to use AI primarily to reduce headcount because in many cases it doesn’t deliver the expected returns.

The goal of productive deployments is to remove repetitive work, improve employee effectiveness and give staff better insights.

Macquarie uses AI internally to augment rather than replace its service staff. Its "tier zero agent" investigates, triages and correlates information before an issue reaches a human employee. It checks knowledge articles, customer history, previous problem-management records and external resources.

“It's about having an efficient workflow so our people have the space in their day to be talking to our customers and that improves the outcome our customers are going to receive,” he told CRN Australia.

AI is also prompting organisations to rethink their broader cloud strategies, with Wallace saying Macquarie is seeing an uptick in cloud repatriation. “Repatriation is a real thing. The days of migrating everything into a single cloud with the view that you’ll transform and optimise later — they've been gone for a long time,” he explained.

“People are starting to realise the benefit they can receive by pulling workloads, like non-transformed workloads, out of a public cloud and back into a private cloud; or in some cases back on premise or into a colocation facility. They're getting significant commercial benefits,” he added.

Sovereignty is also moving beyond a compliance consideration, with organisations increasingly examining where their data resides, who can access it and where it travels.

“Sovereignty is now seriously on the agenda for customers, although it is not yet driving every technology decision,” he added.