If you've been listening to the noise coming out of Silicon Valley or your own company’s boardroom, you might think the world is being rebuilt by AI agents overnight. But before you get too comfortable with that shiny new chatbot your company just paid for, take a look at the math. Over $1.5 trillion in value has been wiped from the books of Software-as-a-Service (SaaS) companies recently. While everyone is busy talking about the next big model, the truth is that most enterprises are playing a very expensive game of make-believe.
Rohit Kedia, a former software architect who now runs Xoriant, suggests that what we're seeing is a massive disconnect between hype and actual output. Xoriant operates as an AI-native digital engineering firm, meaning they get paid to actually build these systems, not just talk about them. He warns that leaders are caught in a storm of disruption. Their existing tools are being made redundant by AI much faster than their slow-moving IT departments can handle.
The opportunity that is being left on the table sits entirely above that layer. This is the 'apply AI' economy. It's the work of taking models that already exist and embedding them into real business operations.
It sounds simple, yet 88% of companies claim to be investing in AI, while only a third have actually managed to get anything meaningful off the ground. Most of the money is going toward licensing models and playing with platform partnerships. This is the 'model layer' trap—buying the shiny car engine but forgetting you don't even have a chassis or wheels to attach it to. Goldman Sachs noted that AI isn't killing software; it's just making it cheaper to write and more powerful to run. If your business hasn't figured that out yet, you're just lighting money on fire.
There are four specific reasons why these projects end up in the corporate graveyard, and they have nothing to do with how smart the AI model is. First, you have the process issue. If you take a broken, inefficient workflow and automate it with AI, you don't get an improved result—you just get a broken result much faster. Second, there's the technology hurdle. You can't run advanced intelligence on top of archaic, fragmented legacy software systems that were built in the 1990s.
They simply weren't designed to talk to each other.
Third, there is a massive skills gap in the workforce. We aren't talking about training staff to write code, but helping them understand how to work alongside these tools in their daily tasks. Often, companies treat this as an afterthought instead of a core pillar of their business strategy. Finally, there's the data problem. Garbage in, garbage out is still the rule of the day.
A 2025 Gartner report found that organizations are expected to abandon 60% of their AI projects soon, mostly because their data isn't ready for prime time.
If you want to stop wasting money, you need a different way to look at your spreadsheet. Think of it in three tiers. Your 'foundation' budget covers the boring stuff: fixing your data, cleaning up your workflows, and training your people. Your 'intelligence deployment' budget is where you stop using AI as a standalone toy and start embedding it directly into the product or service you sell. Finally, 'agentic orchestration' is the high-level work of connecting these tools so they actually talk to one another.
For businesses across Africa, this shift is even more critical. When you don't have the luxury of endless capital to waste on failed experiments, focusing on the basics is the only way to stay competitive. In Lagos, Nairobi, or Johannesburg, companies that prioritize 'applied intelligence' over 'AI theater' will be the ones that actually dominate the market in the next five years. It’s not about the partnership announcement on LinkedIn; it’s about whether you can actually predict your supply chain demand better today than you could yesterday.