The founder of CrossWise InfoTech Limited, as a full member of the Hong Kong Computer Society (HKCS), recently participated in an offline seminar hosted by HKCS titled “Retail Tech Industry Group Seminar – Unlocking AI with Proven Use Cases.”
The topic discussed at the seminar was highly relevant and appealing to SMEs: leveraging AI technology can significantly boost productivity without increasing headcount.
Having previously served as the IT Director of a large enterprise group, the founder of CrossWise InfoTech Limited is well-versed in using IT to enhance operational efficiency. Therefore, we strongly believes that appropriate adoption of AI can bring even greater productivity gains than conventional IT solutions.
However, any decision must be made from a multi-dimensional perspective. For SMEs, the primary consideration has always been — and will always remain:
Cost-effectiveness.
If the investment does not justify the return, then talk about AI implementation or digital transformation becomes meaningless.
Therefore, when potential returns are uncertain, how to effectively control the cost of adopting AI while ensuring the business keeps up with technological trends—without overextending and becoming a casualty—is a key dilemma for business owners.
Deploying AI on-premises requires purchasing and setting up AI all-in-one servers, costing anywhere from tens of thousands to over one hundred thousand Hong Kong dollars. More importantly, businesses must have a suitable environment to house these machines for stable operation. In a space-constrained city like Hong Kong, this is often a major challenge. Fortunately, electricity costs remain relatively low.
On the other hand, cloud-based AI services are typically charged per “token.” Since this is a technical term, and different languages, word choices, sentence structures, and even particles can affect token consumption, CrossWise recommends estimating costs assuming 1 Chinese character equals 1 token.
For standard use cases that do not involve complex computations or analysis, the cost is actually very affordable. Processing one million Chinese characters costs approximately HKD 35 — roughly equivalent to a two-dish lunch set.
Of course, some business owners may wonder: one million Chinese characters sounds like a lot, but how long would that really last?
The answer depends entirely on the specific use case. If AI is used internally for non-text-intensive operations, based on CrossWise’s experience, each transaction typically generates around 5,000 characters of data. One million characters could therefore support around 200 transactions.
From this perspective, the cloud-based AI services clearly offers better value.
However, for text-intensive operations, each transaction may generate at least 20,000 characters — four times more than non-text-heavy processes. In this case, one million characters would only cover around 50 transactions. The good news is that such operations usually command higher fees, helping to offset the costs.
Taking the above calculations into account, along with the challenges associated with on-premises deployment, it becomes clear that for SMEs in Hong Kong, the preferred option for AI implementation remains cloud-based AI services.
Yet, implementing AI involves more than just calculating operating costs. There are also hidden expenses such as process automation reconstruction, integration with existing systems, and more.
Even more overlooked by many business owners is what CrossWise will explore in our next article:
Data compliance and privacy protection.
* Article photo is shot by CrossWise.