Malta’s technology landscape is evolving at speed, with Artificial Intelligence rapidly transitioning from an experimental concept to a mission-critical business tool.

Guiding businesses through this transformation requires both strategic insight and technical expertise. As we begin our chat, Loui Mercieca, Head of Engineering at ICON, a prominent software firm within Malta's tech sector, observes two fundamental approaches local businesses are taking towards AI adoption. The first he describes as “more solution-oriented”, which sees companies leveraging existing research and platforms to solve problems. This approach offers a quicker route to market and, as he notes, “limits risk because you know the tools are out there already and you also know their capabilities,” although it may restrict the scope of solutions provided.

The second path involves genuine innovation, which sees companies create altogether new tools rather than limiting themselves to what already exists. This is inherently riskier, yet the potential rewards are substantial. Loui sees Malta’s specific environment as advantageous here. “Since we’re very small, we tend to work in many areas... we are not highly specialised in one but know about a lot – an important skill when it comes to innovation,” he points out. This breadth of knowledge makes the local talent pool well-suited for developing novel AI applications, he argues.

Faced with these trends and the acknowledged skills gap in AI expertise, ICON’s strategy is firmly rooted in its personnel. While acquiring new talent is always an option, Loui emphasises ICON’s focus on internal development. “Our main strategy is to upskill the people that we have,” he asserts, highlighting the value placed on existing staff. “Our people are very loyal and they put a lot of hard work in. That is something that always needs to be rewarded.”

According to Loui, this hybrid model – combining the upskilling of existing teams with targeted external hiring when needed – ensures ICON can address both immediate client requirements and pursue longer-term product development goals.

Sharing his thoughts on delivering tangible AI value, ICON’s Head of Engineering believes that many organisations grapple with balancing the allure of AI innovation against the practical hurdles of implementation. To combat this, Loui stresses the necessity of drilling down to the core business problem before deploying technology. “We really sit down with clients before we come up with strategies,” he says, highlighting the importance of looking beyond the surface request. This deep understanding prevents deploying solutions that don’t address the fundamental need.

Moreover, ICON utilises prototyping extensively, not merely for development, but as a tool for managing expectations and demonstrating concrete value early on. This practical method is vital for measuring ROI, Loui says, often a complex calculation for AI projects. “If we can't put a number onto the problem itself, we don't even consider it,” he states, unequivocally. By meticulously measuring baseline metrics – and comparing them post-implementation, ICON provides clear, quantifiable evidence of AI’s impact.

And as AI becomes more deeply integrated into business operations, governance becomes critical. To this end, Loui outlines ICON’s three key areas of focus that extend beyond standard functional testing, starting with rogue behaviour detection, which addresses the inherent unpredictability of adaptive learning models. Equally important is proactively managing fairness and mitigating potential biases in AI outputs. “Even if statistically an answer is correct, you still need to be careful about how that output is governed,” Loui warns.

Finally, the third focus is explainability – that is, moving away from AI as an impenetrable ‘black box.’ ICON prioritises understanding why an AI model arrives at a particular conclusion, Loui maintains, affirming, “don’t just give me an answer, but tell me how and why, and what information you used to get there.” This is crucial for building trust, he states, enabling effective debugging, and ensuring accountability.

ICON integrates these principles throughout the development lifecycle, applying data filters, defining clear limits of acceptance, implementing post-deployment monitoring, and tailoring governance strategies based on client risk tolerance and specific industry regulations, such as those in gaming or finance, the Head of Engineering explains. 

Ultimately, he continues, addressing the AI skills gap effectively requires a partnership approach. ICON often begins by providing its own resources to “help set things up”, thereby assisting clients in clearly defining their own requirements. However, the long-term objective is client self-sufficiency.

“We want the client to feel that the solution is theirs. Our aim is to have a client which is as self-sufficient as possible, not just in the solution itself but also in the maintenance,” Loui states, and this involves actively supporting clients in building their internal capabilities or identifying the necessary external resources, ensuring that AI integration becomes sustainable within their organisation.

Looking towards the future of AI, Loui expresses particular enthusiasm for advancements in rogueness detection, stating, “this excites me a lot because we can have confidence about moving and pushing certain boundaries outwards.” He is equally excited about progress in AI explainability, which promises more transparent and trustworthy systems. Beyond improving existing AI paradigms, he sees vast potential in applying AI to fields currently untouched by it, especially those involving complex manual or mechanical processes where efficiency improvements could be transformative.

As we conclude our conversation, it is clear that ICON is poised to continue guiding Maltese businesses through the complexities of AI adoption – by grounding AI initiatives in a clear understanding of business needs, focusing on developing internal talent, implementing robust governance, and maintaining a practical approach to measuring value.