Right now, boardrooms and the executive suite across APAC are buzzing with AI ambition. Promises of cost savings, operational reinvention, and competitive edge have created a kind of techno-gold rush. You know something has hit the zeitgeist when the majority are investing in it, and that is happening with AI: 61% of CEOs globally and 52% in Singapore are instructing their organisations to actively adopt AI agents.
But amid the rush to do something with AI, many leaders are bypassing a more important question: Is the AI you’re implementing actually practical?
This isn’t a rhetorical question. Too often, AI implementation veers into vanity territory, with projects launched not to solve a pressing business need but to satisfy a board directive, investor interest, or simple fear of being left behind. The result? Solutions without substance. Hype without impact. And, ultimately, failure.
While all IT projects have a risk of failure, it’s typically in the 25-30% range. With AI, the numbers are higher: MIT Sloan Management Review reports failure rates of around 70%, while a 2018 Gartner prediction suggested that 85% of AI projects would deliver erroneous outcomes due to issues such as bias in data, algorithms, or management. Clearly, the enthusiasm for AI needs to be better coupled with an understanding of what the AI project needs to achieve.
Cultural cost of AI-for-AI’s-sake
AI is often pitched as a silver bullet, but it’s more accurately a mirror. It reflects the quality of the processes and systems onto which it’s layered. If those foundations are unclear or misaligned, AI doesn’t fix the problem. If anything, it magnifies it.
And that magnification comes at a cost. When AI is introduced without a clear ROI or purpose, it erodes confidence in the organisation’s human knowledge base. Staff begin to distrust not only the tools but also the judgement of those who brought them in. That then becomes a cultural issue.
Every system change influences workplace culture. But AI, by its nature, touches everything: how decisions are made, who makes them, and what’s considered true. If you’re not intentional about what AI is supposed to change, you risk changing things that shouldn’t be touched in the first place.
Three streams of AI use
We categorise AI deployments into three streams:
- AI as assistant: This is the most benign and arguably the most commonly useful right now. Think of AI tools that support daily work, such as drafting documents, generating images, formatting PowerPoint presentations, or summarising data. These tools may not revolutionise your business, but they often improve creativity and impact, delivering speed-to-information and light-touch productivity gains. They’re helpful, non-invasive, and rarely culturally disruptive.
- AI as solution builder: We see organisations building custom AI solutions where no real problem exists. They create fatigue and cynicism. The technology itself may be fine, but the lack of purpose renders it a distraction at best and a reputational risk at worst. Done wrong, these projects dilute confidence in what AI can actually deliver and crowd out space for the things that matter.
- AI as automation engine: Done well, automation can transform operations. From customer service to supply chains, automation has obvious applications. But it’s also the most misunderstood. It’s not a one-and-done. Automation introduces a new layer of work that requires monitoring, compliance, quality assurance, and a human touchpoint to ensure outcomes remain aligned with business needs and ethical expectations. That means automation isn’t a pure cost saving. It requires planning, resourcing, and change management to succeed.
Get the house in order before embracing AI
Here’s the uncomfortable truth: AI is not a fixer, it’s an amplifier. If your processes are broken, AI won’t correct them and you’ll find it all the more difficult to unwind once you’re down the rabbit hole
I follow the principle of process first, AI second. Don’t drop algorithms into chaos; fix the foundations first, ensuring that any automation or intelligence we layer on improves something worth improving. This might sound incremental, but that’s the point.
In an era of tech bravado, incrementalism feels unglamorous, yet it is proven to work. It’s the difference between bolting AI onto a shaky structure and embedding it into a system built to handle it.
In many APAC markets, enterprises already rely heavily on platforms such as Microsoft Business Central, Power Platform, and Azure AI. Success, however, depends less on adding new tools and more on how existing systems are configured and integrated. Leaders should ask:
- Are we solving a real, defined business problem?
- Have we accounted for the added overhead of managing AI?
- Is our data infrastructure mature enough to support this?
- Do we have the right human touchpoints in place?
- Will this solution make people’s work better, or just busier?
If the answer to these questions isn’t clear, it’s better to pause on the AI plans until the answers can be sorted through.
To be clear, this isn’t a call to retreat from AI; quite the opposite. AI is essential to the future of efficient, responsive business. Confidence in AI stems from a realistic perspective of it.
If you really want to lead with AI, you have to start with the business. Then work forward, one useful, intelligent, and culturally aligned step at a time.
