The AI Dilemma for Hong Kong SMEs: It's Not That Employees Can't Learn, It's That Your Old Systems Are Holding You Back

Since generative AI tools like ChatGPT and Midjourney swept the globe, almost every owner and manager of a Hong Kong SME has been pondering the same question: "How can our company apply AI?" You've likely arranged training, encouraged colleagues to try new tools, and even purchased subscriptions for several AI software packages. However, months later, the results seem less than ideal. AI appears to remain at the level of creating a few beautiful images or drafting a few emails, unable to truly integrate into core business processes and enhance productivity.

At this point, a common conclusion emerges: "Is it that our colleagues can't learn? Or are they resistant to change?" But hold on. Before you settle on that conclusion, ask yourself another, more fundamental question: Is the company's IT infrastructure ready for the AI era? In fact, what many Hong Kong SMEs are facing is a deeper, hidden bottleneck. The AI dilemma for Hong Kong SMEs is often not that employees can't learn, but that your legacy systems are "holding you back."

The Wrong Culprit: Why Do We Always Point the Finger at Employees?

Blaming employees for the slow adoption of AI is a simple and seemingly logical explanation. It's easy to observe:

• Surface-level "Resistance": When you ask colleagues to use AI to assist their work, they might offer many reasons why it "won't work," such as "The AI doesn't know our customer details," "The data can't be fed into it," or "It doesn't work with the systems we normally use." These sound like excuses but likely reflect real technical obstacles.
• The Expectation-Reality Gap: Owners see awe-inspiring AI demos and expect one-click solutions to all problems. But when employees try to use them in practice, they find the new tools incompatible with old workflows, unable to "connect" smoothly, leading to natural frustration.

Let's imagine a real scenario: A marketing colleague wants to use AI to analyze customer data for more precise promotion strategies. But customer information is scattered across a decade-old on-premise CRM system, several colleagues' individual Excel spreadsheets, and another accounting software that cannot export data. The AI tool is like a supercar engine, but you're asking it to drive on a rough, muddy road—the result is predictable. This isn't a driver (employee) skill issue; it's a road (system) problem.

The "Four Sins" of Legacy Systems: How They Become AI Stumbling Blocks

Outdated IT systems (Legacy Systems) act like invisible handcuffs, locking down a company's digital transformation potential. Regarding AI application, their main "sins" are the following four:

1. Data Silos: AI's "No Rice for the Pot"

AI's power comes from data. It requires vast amounts of clean, interconnected data for learning and analysis. However, the reality for many Hong Kong SMEs is the prevalence of "data silos." Financial data is in System A, sales records in System B, and customer service logs in System C, all completely independent and unable to communicate. When data is trapped in these silos, AI cannot get a complete picture of the enterprise. It cannot link sales growth to a successful marketing campaign or predict churn risk based on customer complaint history. With data scattered like grains of sand, even the most powerful AI faces "no rice for the pot."

2. Lack of Integration & APIs: A "Chicken Talking to a Duck" Communication Failure

The design philosophy of modern software services (SaaS) is openness and connectivity. Most offer APIs (Application Programming Interfaces), like standardized plugs, allowing different systems to easily exchange data and instructions. AI tools rely on these APIs to "embed" into existing workflows. However, legacy systems, especially those custom-made on-premise ones, are often closed black boxes with no concept of APIs. This means you cannot connect an AI chatbot to your old customer database, nor can you have AI automatically summarize meeting notes and save them to your local file server. The result: employees must constantly copy and paste between old and new systems, reducing efficiency instead of improving it, naturally losing confidence in so-called "AI empowerment."

3. Insufficient Performance & Scalability: A Vintage Car Can't Catch the High-Speed Rail

AI computation, especially the training and running of machine learning models, requires enormous computing resources. That eight-year-old server in the corner of your office, which might be adequate for daily file access and email, is undoubtedly being pushed beyond its limits if asked to also handle AI model calculations. This leads to slow AI tool responses or even system crashes. In contrast, modern cloud platforms (like Microsoft Azure, AWS) offer virtually unlimited, elastic computing power that can scale up or down on demand. Clinging to outdated on-premise hardware simply cannot provide this flexibility and performance, severely limiting the depth and breadth of AI applications.

4. Security Risks & Compliance Issues: An Opened Pandora's Box

Legacy systems often mean end-of-life support, riddled with security vulnerabilities. In the process of introducing AI, you need to transmit company data to AI service providers for processing. If your underlying systems are inherently insecure, this process exposes sensitive company information to significant risk. Furthermore, with regulations like the Personal Data (Privacy) Ordinance (PDPO) becoming increasingly strict, ensuring compliant data use in AI applications is a major challenge. Legacy systems often lack granular permission management and data tracking features, making it difficult for companies to prove the legality of their data processing, potentially breaching regulations at any time.

Breaking Free: Paving a Broad Road for AI

After recognizing the root of the problem, the next step is to take action. Introducing AI to your business cannot be a case of "treating the head when the head hurts, the foot when the foot hurts." It requires a more forward-looking strategy, the core of which is modernizing your IT infrastructure.

Step One: Conduct a Comprehensive "IT Health Check"

First, you need a clear understanding of the current state. Hire professional IT consultants (like Frasertec Limited) to conduct a thorough audit of your company's existing hardware, software, network architecture, and data management practices. Identify the "bottlenecks" slowing progress and which systems urgently need upgrading or replacement. This diagnostic report will form the basis of your future roadmap.

Step Two: Embrace the Cloud, Break Down Data Barriers

This is the most crucial step. Gradually migrate your core business systems—such as Customer Relationship Management (CRM), Enterprise Resource Planning (ERP), and document management—to modern cloud-based solutions (like Microsoft 365, Google Workspace, Salesforce). Moving to the cloud not only enables your team to work from anywhere but, more importantly, integrates previously scattered data onto a unified platform, fundamentally breaking down "data silos."

Step Three: Prioritize SaaS Solutions with Open APIs

When selecting any new software service in the future, besides functionality and price, be sure to consider "whether it offers open APIs" as a core criterion. Software with a healthy API ecosystem means it has limitless expansion potential and can easily integrate with any future AI tools or other systems you wish to introduce, preserving maximum flexibility for your business.

Step Four: Start Small, Build Success Stories

You don't need to replace all systems at once. Adopt a "point-to-area" strategy. Choose the business area most in need of improvement—such as customer service or content marketing—and be the first to introduce cloud tools and AI applications. For example, deploy an AI Chatbot integrated with your cloud CRM to automatically answer common customer questions. When the first project succeeds, employees will personally experience the benefits of new technology, clearing obstacles for broader implementation and building company-wide confidence.

Your Employees Are Ready. Are Your Systems?

In conclusion, the crux of the AI dilemma for Hong Kong SMEs lies not in human capability or willingness, but in the technological foundation supporting business operations. Your employees, especially the younger generation, are already accustomed to various smart applications in daily life. They are eager to bring these efficient tools into the workplace. When they feel frustrated, it's often not because they "can't learn," but because the company's old systems make them "unable to use" them.

Investing in IT infrastructure modernization is not a mere expense but a fundamental investment in your company's future competitiveness. It is paving the way for the coming AI revolution, ensuring your business is not left behind in this wave. Stop blaming your team for failed AI implementation. Instead, examine your systems and ask yourself: Are they empowering your employees or holding them back? It's time to act. Conduct a thorough IT upgrade for your business, unleash your team's true potential, and make AI a powerful engine for your business growth, not an unattainable dream.