Just six months ago, most conversations around artificial intelligence in Malta and the rest of the world centred on experimentation.Businesses were exploring ChatGPT, Microsoft Copilot and Gemini. Employees were using AI to summarise documents, draft emails and automate simple tasks. The dominant question was whether AI could be useful.

Today, according to Finian Massa of ICT Solutions, that question has already been answered.

"The product is genuinely good," he tells WhosWho.mt. "For certain things, it is amazing."

The challenge now is something far more complex: how organisations govern, measure and integrate AI into the fabric of their operations.

"We've seen two or three massive leaps in AI models since we last spoke," Mr Massa explains. "Companies are now far more keen to adopt these tools because they can see tangible value."

What has changed, he argues, is that AI has evolved beyond being a sophisticated chatbot.

"This morning, I sent out around 35 emails," he says. "They weren't sent by me; they were quickly reviewed by me. The system drafted them, checked branding, identified the right recipients and prepared them for sending."

Similarly, AI is increasingly being used to conduct research, analyse documents, compare information across multiple sources and even assist in website development and content strategy.

"There is now a level of autonomy that is taking off," he says.

Yet while the productivity gains are becoming obvious, so too are the risks.

Governance arrived sooner than expected

One of the most striking admissions from Mr Massa is that his previous predictions around AI governance proved wrong.

"I thought governance would not really become a conversation until 2027," he says. "Hand on heart, I was wrong."

The reason is simple: AI capabilities have advanced far faster than expected.

As businesses move from experimentation to deployment, questions around security, privacy and accountability become unavoidable.

"There is a tension," he explains. "On the one hand, to be effective as a business, you need to be using the right tools. On the other hand, directors and management teams want to feel comfortable that they are not exposing the organisation to unnecessary risk."

This challenge is particularly evident in regulated industries.

Mr Massa points to concerns around data residency, especially where advanced AI models may process information outside European jurisdictions.

"Businesses are asking: can we justify the increase in risk to get the new enablement tools?" he says.

For many organisations, the answer is increasingly becoming yes, but only with proper controls, stronger data hygiene and clear governance frameworks.

"You cannot fall behind the cutting edge simply because you're trying to be holier than the Pope," he remarks.

The difficult question: How do you measure AI?

Perhaps the biggest challenge facing executives is proving return on investment.

Traditional automation projects are relatively easy to evaluate. If a chatbot reduces customer support calls by 80 per cent, the value is immediately visible. Knowledge work is far more complicated.

"You cannot measure an engineer by the number of lines of code written," Mr Massa says. "You cannot measure a salesperson by the number of emails sent. Those are dumb metrics."

Instead, businesses are wrestling with a more difficult question: whether AI is genuinely making employees more effective.

An employee may still work an eight-hour day, but produce significantly more output, make better decisions or spend less time on repetitive tasks.

"Whether before AI or after AI, I still have eight hours of productive work," he explains. "The challenge is understanding what has changed within those eight hours."

The answer, he believes, will require a combination of usage data, business outcomes and employee feedback.

"Just because somebody is using the tool doesn't mean they're more effective," he says. "And just because they're not using it constantly doesn't mean they're not getting value."

Why taste may become more valuable than intelligence

One of the more provocative ideas emerging from Mr Massa's thinking is that AI may fundamentally change which skills matter most.

Historically, access to information and technical expertise created competitive advantage.

Now, those advantages are increasingly available through AI systems.

"I think the most important thing is becoming taste," he says.

The ability to recognise quality, identify weaknesses and direct AI towards better outcomes may become more valuable than producing the first draft itself.

"You need to know enough about the subject to say this is right and that is wrong," he explains.

"You need to know enough about design to say, hold on a second, don't make the PowerPoint like that. Position it differently. You need to know enough about storytelling to know when the narrative isn't working."

In this world, AI becomes less of a replacement for expertise and more of an amplifier for people who already possess judgement.

"If you have the taste to direct, you can do so much more," he says.

The next battleground isn't AI models. It's infrastructure.

Looking ahead, Mr Massa believes much of the public conversation is focused on the wrong thing.

He references Amara's Law, the idea that society tends to overestimate the short-term impact of technology while underestimating its long-term consequences.

The historical examples are familiar.

The lightbulb existed before electrical grids. Shipping containers existed before ports and logistics systems adapted to support them.

For Mr Massa, AI finds itself at a similar moment.

"We have the product," he says. "The question is what does the infrastructure around it need to look like?"

That infrastructure includes governance, compliance frameworks, integration layers and access to computing power.

His concern is that Europe may be falling behind in one particularly important area: sovereign computing capacity.

"We don't have enough compute," he argues.

While AI companies continue to race ahead, Europe remains heavily dependent on infrastructure controlled elsewhere.

"If the United States has the data centres and China has the data centres, but we don't, how are we going to run these models ourselves?" he asks.

The issue, he believes, extends beyond business competitiveness.

"It starts to become a question of strategic importance."

Are AI companies overvalued?

Mr Massa also questions whether today's AI market leaders will ultimately capture the value many investors expect.

"The models are overvalued and the infrastructure is undervalued," he argues.

While companies such as OpenAI, Anthropic and Google continue to release increasingly powerful models, competitors rapidly close the gap.

"Every time somebody hits a new benchmark, another company shows up a few weeks later and reaches something very similar," he says.

As a result, he believes the real strategic advantage may not lie in the models themselves but in the computing infrastructure that powers them.

"What becomes strategic is whether you have the capacity to run these systems."

The result may be a future where AI models become increasingly commoditised while access to large-scale computing resources becomes the real competitive differentiator.

A strange but exciting moment

Despite the uncertainty, Mr Massa remains optimistic.

He describes the current period as both exciting and unsettling.

On one hand, AI is enabling levels of productivity that would have been difficult to imagine even eighteen months ago.

On the other, businesses and individuals are still working out how much responsibility they should delegate to machines.

"It's a very interesting, very weird time," he says.

Yet if there is one conclusion emerging from the rapid evolution of artificial intelligence, it is that the conversation has fundamentally changed.

The debate is no longer whether AI works.

The debate is whether organisations can build the governance, infrastructure and expertise required to use it responsibly.

And, according to Finian Massa, that conversation is arriving much sooner than anyone expected.