About
Subscribe

Can the cloud's missteps help us deploy better AI?

We can harness the cloud era's lessons to develop and prove AI's value faster.
Johannesburg, 19 Nov 2026
Dale Slabbert, CEO at TrueMark South Africa. (Image: AI-enhanced)
Dale Slabbert, CEO at TrueMark South Africa. (Image: AI-enhanced)

The cloud was a seminal moment in Dale Slabbert's career. When he worked at Adobe, cloud technologies stepped onto the scene, offering substantial technology leaps for software engineers and other technical professionals. The transition also brought about another very substantial change.

"Using the cloud sped up how we did things, but it also sped up the spend," said Slabbert, who today is CEO at TrueMark South Africa. "At Adobe, when that happened, we had to create a new team just to manage that spend."

Mounting cloud costs were one of the new challenges; another was on how to understand and exploit the new technology opportunities. After an initial fervour for "lift-and-shift" migrations that divorced crucial software from taxing hardware, Slabbert realised that there is much more the cloud can offer once you appreciate that it does things differently.

Learning those lessons was expensive, both in rising bills and exploring avenues that went nowhere. As companies take hold of the AI era, they risk repeating similar mistakes. But if they approach things smartly, they can avoid them.

Echoes of cloud follies

Adopting AI creates similar costly fallacies and conjectures, echoing the uncertain cloud years. Slabbert points to a common mistake: the assumption that AI is always the best form of automation.

"We've built probably 40 different AI POCs this year that people have approached us on. Of those 40 different POCs, around 10 actually hit the ground running. Many people simply want to automate everything with AI, but it's not always the right tool for that situation. In a couple of instances, probably 40% of the cases, we ended up coding the automation because AI was more expensive, slower and less accurate."

These experiences help explain a stubbornly high failure rate among companies to bring AI into production. Around 80% of AI projects still fail (RAND), and only 37% of organisations report a positive impact on earnings (McKinsey), the latter remaining unchanged from previous surveys.

Companies are seeing benefits from AI, but much like the early cloud era, they are stumbling upon them and often paying a lot in the process.

Learning from the cloud era's lessons

Today's cloud landscape is very different. Companies have learned to be more selective, calibrating cloud adoption on a case-by-case approach and leaning on the experiences of skilled partners to get the best results.

AI projects benefit when they use this same mindset. Some tasks and workflows are very suited to automation, and some automation could replace human involvement. But there are also other types of automation that don't need AI, and there are clear examples where AI works best when it supports people rather than replacing them.

"Not everything should be built into that agentic AI space; that's kind of the mentality that our own learnings, failures and growth opportunities have created," says Slabbert, reflecting on TrueMark's high concentration of coding and software engineering expertise.

Re-imagine instead of retread

After being wowed by AI, he resisted the temptation to start implementing it everywhere. Instead, he gave his engineers budgets and challenged them to come up with use cases. That approach helped solidify their understanding of AI's potential and pitfalls, reflected against their established experiences in developing advanced AWS solutions.

"We are able to take those experiences back to customers and say, 'We built this for you like you asked, but we don't think this is the best solution for you. So here's where we would go from here.' Then we build the actual automation and code for them. We re-imagine what could be done with AI."

Much like early cloud adoption, there is a temptation to treat AI as simply a better way to do what was done before. Yet, that philosophy leads to higher costs and underwhelming outcomes. Instead, Slabbert and his team at TrueMark know that success materialises when we let go of AI dogmas and focus on the outcomes.

"There are a lot of preconceived notions of what AI is, and this fear that if companies don't act quickly, they'll get left behind. But that takes you down the wrong path. The cloud era taught us that. It also taught us that if we take the time to understand the technology, focus on what we want to create and work with the right people to build what works, the results are incredible."

Share