Junto Innovation Hub
IT Get in touch →

Magazine Artificial Intelligence

Quando l'intelligenza artificiale sostiene reti e competenze

3 min read

Quando l'intelligenza artificiale sostiene reti e competenze

The adoption of intelligenza artificiale is changing infrastructure, skills and products in companies. The piece brings together practical examples from the electric sector, data center development, open source and digital insurance. The choices traced are useful to founders and investors. Here it describes how GPUs, real data, technical translation and loyalty programs intertwine and which regulatory and operational constraints emerge.

When intelligenza artificiale supports networks and skills

Regional electric companies use language models and computer vision to automate inspections and maintenance in complex territories. One system analyzed 26,000 drone images and found over 3,600 defects, a figure that shows the possible scale jump in diagnostics. The same organizations are building multilingual corpora to cover Laos, Vietnam and Malaysia and are developing translation tools with terminology for more than 1,800 types of electrical equipment, because technical communication remains critical in cross-border activities.

Investments in human capital are also under way. Two cohorts totaling 53 students have entered joint programs between companies and universities for practical training. Training local talent on real cases is an integral part of the implementation strategy. >Automated technical translation and shared labs reduce operational risks and accelerate the spread of best practices. >Working on real data speeds up model reliability in the operational context.

Building compute capacity and choosing between open and closed

Compute capacity has become a strategic lever for digital sovereignty and for access to specialist AI products. A Southeast Asian company will bring its “AI factory” beyond 100 megawatts and will deploy 22,000 GPUs across two facilities in Malaysia and the Philippines at the beginning of 2027, reaching over 48,000 GPUs in total. More GPUs means access to advanced compute for startups, companies and local institutions.

The European debate pushes toward open source on the model front. The choice serves to avoid dependencies on large foreign providers and to maintain control over data and infrastructure. The operational numbers are clear. Up to 88% of AI pilot projects stall before production because of governance, data and infrastructure costs. General models reach about 70% effectiveness on specialist tasks while teams seek performance above 90%. Training open specialized models on company data is the path indicated to reach that level. >Using open weights can reduce inference costs by up to 87% compared with closed solutions.

This choice also has a regulatory constraint. At the European level the AI Act requires transparency and metadata. Californian regulation imposes disclosure obligations from 2027 for some platforms and changes the relative value of opening or closing weights.

Digital customer journey, loyalty and product opportunities

Digital experiences are the node that connects infrastructure and models with the market. A healthcare operator in Hong Kong recorded a 38% increase in visits to its online subscription platforms, a figure that signals growing demand for simple and comparative tools. That platform allows instant quotes, side-by-side comparison and support via WhatsApp; it allows saving progress and paying with QR code or local systems. Simplifying the digital journey increases customer trust and conversion.

On the product side, a Singapore insurer launched Travel Pass with four tiers. Bronze is obtained with S$100 of spending and Platinum with S$1,000. Benefits include global eSIM data, vouchers and cashback without claims; Gold and Platinum also offer additional personal accident coverages. Tying rewards and protection incentivizes loyalty and turns insurance into a recurring service. Current promotional offers include discounts up to 55% on some policies until 28 December 2026, a factor that can accelerate initial adoption.

Predictive models and personalization enabled by GPUs and open models can power recommendations, detect fraud and calibrate premiums dynamically. However, data quality and governance remain binding conditions for relying on these automatisms.

Strategic trade-offs for founders and investors

The synergy between infrastructure, open models and customer experience creates opportunities but requires deliberate choices. Investing in local compute requires governance on energy and sustainability, while betting on closed models can offer speed but increases the risk of lock-in. Security improves with transparency and local control of data, a principle that pushes toward open architectures and preventive compliance.

Founders and investors must evaluate three levers together: local compute capacity, specialist models trained on proprietary data and a digital customer journey that converts and retains. Public-private partnerships, joint labs and practical training programs reduce operational risk and create a pool of local skills. The main challenge remains open. GPUs, data and models must be transformed into regulated and scalable products. Compliance must become a competitive advantage, not an obstacle. The real difference will be measured in the ability to turn data and GPUs into useful and regulated services.

Sources:

Related articles