South African companies are spending heavily on AI projects, but the real bill is often much bigger than the one that gets approved in the boardroom. According to Jason Harrison, chief operating officer of The Up & Up Group, many executives are rushing into the hype without fully understanding the hidden costs of AI, from governance and integration to legal exposure and runaway cloud usage.
Harrison’s warning lands at a time when local businesses are under pressure to modernise quickly. In sectors ranging from banking and retail to logistics and professional services, AI is being sold as the answer to everything: faster workflows, leaner operations and smarter decision-making. But as we’ve seen across the market, technology projects rarely fail because of ambition alone. They fail because the real work begins after the announcement.
Speaking on TechCentral’s TCS+, Harrison drew on The Up & Up Group’s own experience of testing and implementing AI to show how the gap between promise and reality can widen fast. He argues that many organisations focus on the shiny front-end features, while ignoring the harder, less glamorous part of the process: making AI work inside existing systems, policies and teams.
That integration layer, he says, is where a lot of projects stall. It is also where costs start to multiply. Businesses often budget for software licences, model access and some technical rollout support, but miss the longer tail of spend that comes with operationalising AI properly. In practice, that means technical adaptation, staff training, oversight, ongoing tuning and the management structures needed to keep systems under control.
One of Harrison’s key points is that policy and governance should never be treated as side issues. In many organisations, they are seen as paperwork or compliance overhead. But in the AI era, they are becoming core expenses. Companies need rules for what employees can feed into models, how outputs are checked, who signs off on use cases and what happens when something goes wrong.
There is also the question of risk, and that is where the hidden cost of AI becomes even more serious. Harrison warned that if copyrighted material appears in outputs, or if an autonomous agent behaves outside its intended limits, the fallout can extend well beyond embarrassment. Businesses can face reputational damage, financial exposure and even legal consequences. Those are not theoretical risks for South African firms operating in a tight regulatory and consumer environment.
The hidden cost of AI is not only about money, either. Harrison also pointed to the broader energy footprint of the technology, saying the industry is still struggling to come to terms with the amount of power and infrastructure needed to support heavy AI use. As companies scale their deployments, they are effectively adding to an environmental burden that is often left out of the conversation.
Why the hidden cost of AI is becoming a boardroom issue
For South African executives, the challenge is that AI adoption is accelerating even as the risks become clearer. Harrison described the current moment as resembling a nuclear arms race, where companies feel forced to move quickly because competitors are moving first. In that environment, the fear of being left behind can be as powerful as any formal strategy.
That pressure, he suggests, means much of today’s AI spending is driven by Fomo — fear of missing out — rather than by careful business planning. It is a sobering observation, especially for boards that are under pressure to show innovation, growth and digital transformation in the same breath. The danger is that organisations confuse speed with strategy and end up paying for tools they cannot properly absorb.
Harrison’s advice is to take a more measured route. Instead of rolling out a huge AI programme across the business, he supports a “test and learn” approach. That means running multiple small, low-cost experiments across different teams and use cases, then scaling only what proves genuinely valuable. It is a more disciplined way to spend, and in many cases, a more realistic one.
He also pushes back against one-size-fits-all deployments. Not every department needs the same AI tools, and not every problem needs automation. In some cases, a narrow use case with a clear return will outperform a broad, expensive rollout that looks impressive on paper but struggles in practice.
For Harrison, governance should sit close to the work itself, not only in the executive suite. That is an important lesson for South African companies, where decision-making can still become overly centralised. The people building and using AI need clear guidance, but they also need room to test ideas responsibly. Without that balance, organisations risk creating either chaos or paralysis.
He also believes that short-term targets still matter, but only when they are anchored to a longer-term direction. Quarter-to-quarter metrics can help businesses stay accountable, but they should not push companies into making rushed decisions that compromise strategy. In other words, AI should serve the business model, not the other way around.
Even with his warnings, Harrison is not anti-AI. In fact, he remains optimistic about the technology’s future, particularly on the African continent. His view is that while developed markets may use AI mainly to get answers faster, African organisations could use it to develop deeper thinking and problem-solving ability. That is a subtle point, but an important one for a region still navigating skills gaps, uneven infrastructure and the need for inclusive growth.
For South African leaders, that makes the hidden cost of AI more than a technical issue. It is a strategic, financial and social question all at once. The companies that will benefit most are unlikely to be the ones that spend the most or move the fastest. They are more likely to be the ones that understand what AI really costs, build the right guardrails and scale with discipline.
As we reported earlier, the debate around AI in South Africa is shifting from excitement to execution. Harrison’s message is clear: the opportunity is real, but so are the bills that come with it. For now, the smartest move for many businesses may be to slow down, test properly and make sure the promise of AI does not outpace the reality of what it takes to deliver.