There has been so much buzz around AI in the past few months that it’s hard to separate what AI can actually do from what many people think it can do.
AI is becoming reminiscent of the dot-com boom of the late 90’s—if you aren’t part of it, you’ll supposedly be left behind and regretting your life choices. But AI remains such a broad umbrella term, and the widespread misunderstanding around what it actually is creates short-term implementation problems for many businesses rushing to adopt it.
AI is not a silver bullet
First, it’s Artificial Intelligence. This means it can draw from existing knowledge sources and training data, allowing it to recognise patterns and compile information in new ways. But it doesn’t truly create NEW ideas—it generates outputs based on statistical patterns in its training data.
I hear it all the time: “Just get AI to do it.” Yes, great idea. What AI system? Which Large Language Model (LLM)? What foundation model, fine-tuning, workflow, or automation are you talking about? What integrations have you set up to capture relevant, usable, and actionable results?
You can’t just ask AI to do whatever you want unless you understand what the various models’ capabilities and limitations are, how to prompt them effectively (prompt engineering), and what you’re ultimately hoping to achieve. And even then, every AI system remains limited in fundamental ways (for now).
The power is in the prompt
The power of utilising AI effectively lies in how you prompt it. You have to understand how to craft inputs that extract the result you want and let the model do the appropriate heavy lifting.
If you ask ChatGPT to “build a website for my flower shop,” it may ask a few follow-up questions and then generate some basic HTML code. Now what? Are you going to learn how to deploy that code, add functionality, connect the CMS and maintain the site?
If you ask Claude to write you the outline of a marketing proposal, it will do a decent job giving you something that resembles one. You can certainly edit this and make it your own—but the initial output quality depends entirely on your prompt quality.
The difference in AI utility comes down to the complexity gap between what you want to achieve—whether you need simple, time-saving content generation or complicated, specialised technical outputs requiring domain expertise.
There will likely be a time when AI interfaces become more intuitive and capable of understanding vague instructions, but we’re not there yet. Effective prompting remains a technical skill.
Real costs vs. promised returns
The narrative that AI will instantly cut costs and boost productivity is pervasive but oversimplified. What many vendors won’t tell you are the hidden costs of implementation. There are the subscription fees for enterprise-grade models, sure—but what about the total cost of ownership including time spent learning systems, developing prompt libraries, integrating with existing infrastructure, and implementing human review processes?
For every hour “saved” by AI, many businesses spend significant time on troubleshooting, retraining staff, and correcting AI-generated errors (hallucinations). The integration isn’t seamless. Your data might need extensive cleaning and standardisation before any AI system can work with it effectively. Your business processes might need complete overhauls to accommodate AI workflows and establish appropriate guardrails.
The promised 10x productivity boost often becomes a much more modest improvement after accounting for all the adjustments needed and implementation overhead. Is this still valuable? Absolutely. But it’s nowhere near the revolution being promised in breathless marketing materials and vendor presentations.
Humans aren’t being replaced.
Despite what some AI evangelists claim, humans remain essential to meaningful work. AI can generate content but lacks true creativity and original thinking. It can analyse data patterns but lacks contextual understanding of your specific business environment and industry nuances. It can automate routine processes but cannot build authentic relationships with clients and partners.
Most importantly, AI cannot make ethical judgements or strategic decisions about what’s right for your business. It can present options based on historical patterns, but the critical decisions still rest with human leadership. The businesses successfully implementing AI aren’t replacing their workforce through automation—they’re augmenting human capabilities, creating human-AI collaborative workflows that free people from mundane tasks to focus on uniquely human contributions like innovation, empathy, and strategic thinking.
Are you adopting or avoiding?
AI is here to stay. It will increasingly permeate through all aspects of business operations. Better to develop competency now than be left scrambling to catch up. You can dabble, find experimental uses for it, or you can use it strategically to transform your business processes, improve your service offering, and increase operational efficiencies.
The question isn’t if your industry will be affected, but when—when will you develop the AI literacy (at least to an intermediate level) to understand which tools and models will genuinely improve business outcomes, not just create distractions or superficial “AI washing” of existing processes?
If you avoid it entirely, thinking it won’t impact your sector… that’s undeniably naïve. But rushing in without a coherent implementation strategy is equally foolish. The businesses that will thrive aren’t those embracing AI uncritically or avoiding it entirely, but those approaching it with responsible innovation and clear-eyed pragmatism about what specific AI capabilities can and cannot do.
In the end, AI is just another set of business tools—neither magical solution nor existential menace. Like any technology, its value depends entirely on how well you understand both its capabilities and its limitations. The competitive advantage is real, but only for those willing to separate the hype from the reality and build genuine AI competency within their organisations.
