Technology spend needs to move from consumption to accountability, writes Brandon Moodley, chief financial officer at Cloud On Demand.
I do not need a utilisation report. Tell me what changed in the business because we spent the money.
Too many cloud conversations still begin with infrastructure, licences and consumption. That is backwards. They should begin with the outcome or the risk we are trying to manage.
That distinction matters because technology is a capital-allocation decision. A cheaper cloud bill can still be a poor investment, just as a larger one can create real value. The question is not whether we spent less, but whether every rand bought us more capability, resilience or reduced risk.
Recently, a financial services organisation spent around $50 000 collecting and storing AI data before a single customer query had been answered. Nobody had done anything wrong technically. Nothing needs to fail technically for the investment to fail commercially.
This is the trap most cloud and AI conversations fall into. We measure what we can see easily: spend, tokens, consumption, and call it accountability. It is not. A utilisation report tells me what happened last month, not what problem was solved, for whom, or what it was worth. Finance needs the other half: what did the organisation get in return?
From where I sit, finance teams should be asking three simple questions.
First, what is the unit of business value? Before approving the technology, we should understand what we expect to improve and how we will recognise success. Not every workload needs to generate revenue directly. Some exist to reduce risk or build capabilities we could not build ourselves. But every investment needs an economic hypothesis.
Second, what is our full economic exposure? Many organisations running AI workloads through a hyperscaler cannot answer a simple question: which jurisdiction holds our data, and what happens if that changes. Under POPIA, that movement needs to be understood and protected, not treated as a technical footnote. It belongs in the risk architecture of the deal, not just the financial one.
Third, are we paying for compute, or for problems solved? Consumption billing suits infrastructure far better than it suits intelligence. Finance does not care how many tokens were processed, it cares whether the AI resolved more queries, sped up decisions or lifted productivity. That is the shift partners need to price for.
If the only thing a partner can tell me is how much cloud I consumed, they have not justified their place in the conversation. Good partners explain what that spend achieved.
South African businesses are under real pressure to get this right. The combination of a volatile rand, dollar-denominated cloud spend and rising local infrastructure costs from AI demand means approving cloud spend on faith is no longer affordable. ITWeb reported in March 2026* that South African firms should brace for continued pressure on cloud and hosting costs as that demand plays out locally.
None of this is a request for cheaper cloud. It is a request for a commercial model that catches up with how the business actually runs, on problems solved, not on what was consumed. Tell me what we spent, but also what got faster, safer or more profitable, and whether to scale it, change it or stop it. That is the cloud conversation worth having, and a technology bill I can defend to a board, not because it is cheap, but because it is accountable.