Why most AI projects fail after launch

South African businesses are under growing pressure to adopt artificial intelligence (AI), but according to digital transformation consultancy DY|DX, many organisations are focusing on the wrong part of the challenge.

While much of the market conversation remains centred on tools, platforms and automation, DY|DX argues the real barrier to successful AI adoption is far more human: organisational behaviour, workflow friction and the reality of how people actually work inside companies.

According to Nevo Hadas, founder and partner at DY|DX, businesses are often treating AI as a technology problem when it is fundamentally a transformation problem. “The technology is no longer the hardest part. AI tools are becoming more affordable, more accessible and easier to implement every month. The real challenge is redesigning the way people work and making new systems something teams actually want to use.”

The consultancy says global thinking is increasingly validating what it has experienced across years of work in sectors ranging from telecoms and media to retail and emerging markets.

One example is the growing adoption of the 70-20-10 transformation model, which suggests successful AI implementation depends far more on organisational change and capability building than on technology itself.

Under the framework:

• 70% of success comes from organisational and behavioural change.
• 20% comes from skills and capability development.
• Only 10% comes from the technology itself.

“For years, companies approached digital transformation with an ‘if we build it, they will come’ mindset,” Hadas explains. “But we’ve now entered a very different era. Businesses are discovering you can build powerful systems that nobody uses, work around or that people actively resist.”

He says one of the biggest mistakes organisations make is applying AI to broken workflows and existing inefficiencies without first understanding the deeper organisational dynamics behind them.

“Many businesses are simply creating faster dysfunction. AI layered onto a poor process doesn’t solve the problem. It often amplifies it.”

DY|DX refers to this challenge as the ‘hidden organisation’: the gap between how businesses believe work happens and how work actually happens in practice.

“Every organisation has two operating systems,” Hadas explains. “There’s the official workflow on paper, and then there’s the reality of how people navigate approvals, relationships, delays, incentives and interpersonal dynamics every day. If you don’t understand that hidden organisation, transformation initiatives struggle to gain traction. This is true even in large organisations with thousands of employees.”

He says businesses often misdiagnose why tasks take time or where inefficiencies truly exist. “Things don’t always take long because they’re difficult. Often they take long because people avoid them, don’t trust the process, or because there’s friction between teams or individuals. That’s the kind of organisational reality many AI and automation conversations completely ignore.”

According to DY|DX, successful transformation increasingly depends on understanding not only workflows and systems, but also human emotion, trust and behavioural design.

“One of the biggest barriers to AI adoption is not learning new tools, but unlearning old ways of working,” Hadas says. “Businesses often underestimate how deeply existing habits, workflows and organisational norms are embedded into daily operations.”

He adds that this is why organisations need to move beyond viewing AI purely as a cost-saving or efficiency tool. “The real opportunity is not simply doing the same work faster,” he explains. “It’s redesigning workflows so teams can produce materially better outcomes within the same amount of effort.”

Tasks that once produced a single static output can now generate richer analysis, multiple delivery formats, simulations, reporting layers or client-ready materials within the same operational effort.

“That’s where the real value starts emerging – not just reducing effort but dramatically improving the quality and sophistication of what teams can produce.”

As organisations continue rushing to deploy AI solutions, DY|DX believes the businesses that will ultimately succeed will not necessarily be the ones investing in the most tools, but the ones redesigning work most intelligently.

“AI will not replace organisations that move slowly. Organisations that cannot unlearn will replace themselves,” Hadas concludes.