AI without connectivity: 2 billion people left out of the game

AI doesn't arrive on its own: first you have to be connected

The International Telecommunication Union (ITU) has put figures to a reality that goes unnoticed in the day-to-day life of a Spanish SME: around 2 billion people worldwide still lack Internet access, a quarter of the population. The figure coexists with global progress —the organization now counts 6.1 billion connected users— but it makes clear that the expansion of connectivity is not homogeneous, let alone universal.

The ITU's warning is direct: in the midst of the artificial intelligence era, being left out of the Web means being excluded from educational, employment and economic opportunities. Its secretary-general, Doreen Bogdan-Martin, sums it up when she states that "the expansion of connectivity must be meaningful for everyone." It is not just about having coverage, but about that connection serving a purpose.

Two technological speeds

The report paints a map of contrasts that is worth reading carefully. In high-income countries, Internet access reaches 93% of the population; in low-income countries, barely 32%. The gap is not measured only in users: 5G coverage reaches 86% of the population in the most developed economies and stays at 4% in low-income countries. Average monthly mobile data traffic per subscription speaks for itself: 20 GB in rich countries versus 3.4 GB in the poorest.

There are more axes of inequality. Globally, 79% of men use the Internet compared with 74% of women; among young people aged 15 to 24, penetration rises to 86%, twelve points above the rest of the population; and when comparing city and countryside, 82% of urban dwellers are connected versus 63% of rural ones.

Price remains a wall

Affordability is improving, but not enough. The global average price of a basic mobile broadband basket represents 1.2% of gross national income per capita, and fixed access stands at 2.3%. However, in the least developed countries mobile access is equivalent to 6.1% of per capita income. In addition, three out of every ten economies in the world still do not meet international affordability targets for either fixed or mobile broadband.

The ITU insists that access, its quality and digital training will determine who can benefit from the AI boom. Cosmas Luckyson Zavazava, director of the Telecommunication Development Bureau, frames it as a two-way task: it is not enough to expand infrastructure, digital skills must also be promoted. The risk, the organization concludes, is that AI and other emerging technologies will reinforce inequalities instead of reducing them.

What this means for an SME

At first glance, the report seems like a matter of international policy. But it has a very concrete reading for any IT team. The first: AI is not a shortcut to skip the fundamentals. If a company does not have stable connectivity, adequate devices and trained staff, adopting generative AI tools will not bring it up to date; it will only add a layer of complexity on a fragile base. The second: the skills gap also exists within organizations, not only between countries. Within the same workforce, there are those who use AI assistants with ease and those who do not know where to start.

The third reading is from a vendor's perspective. When choosing technology for an SME, it is worth asking whether the solution assumes perfect connectivity that the client does not have. Tools that work locally, that tolerate outages or that degrade gracefully are more valuable than they seem. In this blog we have already seen how hybrid AI, with agents running on the PC itself, helps contain costs and dependencies; and how AI code review requires verifying agents and leaving decisions in human hands. Both ideas fit here: technology must adapt to the context, not the other way around.

Practical recommendations

1. Audit your base before scaling. Review real connectivity, devices and access for your staff. Without that diagnosis, any AI plan starts from the wrong picture.
2. Train before deploying. An AI pilot without prior digital skills generates frustration and abandonment. Dedicate time to basic literacy and specific use cases.
3. Design for intermittency. Prioritize tools that work with modest connections or that allow local work, as we proposed when discussing hybrid AI and agents on the PC.
4. Measure total cost, not just the fee. Affordability is not only the price of the service: it includes devices, training and adaptation time.
5. Don't confuse coverage with capacity. Having 5G does not guarantee making the most of AI; what makes the difference is knowing what to do with it and with what data.

The report's conclusion is uncomfortable but useful: without closing the access and skills gap, AI will widen differences instead of reducing them. For an SME, the translation is simple: before buying the trendy tool, make sure your team and your infrastructure can sustain it. As we already analyzed when discussing the choice of AI models, the decision is not only technical: it is about context.

Source: Computerworld. Analysis and adaptation: ForgeNEX.