This summer, while on vacation and working in Spain, a headline in a Basque newspaper caught my attention: 'Three out of every 10 Artificial Intelligence projects in Spanish companies have been canceled because they end up failing.' The main reason: not knowing how to apply it.
That article made me wonder if an equivalent data point existed for Mexico. I didn't find an official figure, but in my experience as a consultant, I suspect the percentage could be even higher. Here, the dominant narrative is that "we are adopting AI faster than ever," but rarely do we talk about how many projects truly consolidate and generate value.
The growth of AI usage in Mexico is undeniable. According to the report, "The Age of AI in Mexico, Panorama, Trends, and Data 2024," prepared by Banco Santander and Endeavor, the number of AI companies in the country increased by 965% in the last 6 years, reaching 362 active companies and over 11,000 specialized jobs. This boom has attracted more than $500 million in investment, positioning Mexico as a regional benchmark.
But when we talk about impact, the figures take on a different tone. According to the Unlocking AI’s Potential in Mexico study by AWS:
-
72% of Mexican companies that have adopted AI remain at a basic level of use, focused on streamlining processes or routine tasks with public chatbots, without innovating products or services.
-
16% have advanced to an intermediate level, integrating AI more broadly and improving the customer experience.
-
But only 7% have reached the most transformative stage, creating custom AI systems for complex tasks that redefine their operations.
In other words: The majority of current AI efforts in Mexico are not designed to reinvent business models, but rather to make marginal adjustments to what already exists.
As an innovation consultant, I have detected three recurrent obstacles that prevent AI from going from a promise of change to a real engine of transformation:
1. Fragmented Vision
AI often enters as an isolated project, often driven by an IT area or an executive who wants to "innovate" without a cross-functional strategy. The result: disconnected solutions that do not integrate with other key processes.
2. Talent and Training Deficit
The lack of specialists is evident, but the deeper problem is the absence of practical training for the teams that will operate these tools. The assumption is that the technology "speaks for itself," when in reality it demands new frameworks, metrics, and skills.
3. Distrust of Data
AI is only as good as the quality of the data that feeds it. Many companies in our country still rely on manual processes or fragmented systems that generate incomplete or unreliable information. This causes projects to stall or results to be questioned.
In the last two years, I have seen Mexican companies invest significant sums in AI tools for customer service, only to abandon them months later because the models did not fully understand the requests or because the integration with their internal systems was never completed.
In other cases, financial sector companies launch pilots for predictive risk analysis but cancel them upon discovering that the historical database has too many gaps to train a reliable model.
These failures are not always publicized, but they occur frequently and generate a collateral effect: internal skepticism toward AI, which causes future projects to face more resistance to receiving budget or support.
A Challenge More Cultural Than Technical
It's easy to think the problem is only a matter of budget or lack of experts, but the reality is deeper: the true challenge is cultural.
In many Mexican organizations, AI is still seen as an experiment or a "tool to have" to avoid falling behind, rather than a structural change that requires redesigning processes, roles, and ways of measuring success.
In AI terms, we talk about two stages: Deploy (implementing for productivity) and Reshape (redesigning work from the ground up). Mexico, for the most part, remains in the first stage, with notable exceptions in sectors like fintech, logistics, and advanced manufacturing.
If we truly want the use of AI in Mexico to go from an optimistic headline to an engine of competitiveness, there are urgent steps: It is essential to train at all levels, not just leaders. Adoption will be slower if half the organization does not understand how to use the tool or perceives it as a threat. Additionally, AI must be integrated into the central business strategy, not as a parallel project. All while establishing clear impact metrics, going beyond how many users interact with the tool. And finally, ensure data quality, investing in cleaning, integration, and governance.
The reality is that we cannot yet speak of AI maturity in Mexico. Yes, the ecosystem is growing, startups are raising capital, and corporations announce pilots every week. But as long as most projects remain in experimental mode and fail to generate structural changes within companies, AI will be more of an accessory than a pillar of competitiveness.
International experience—including the warning I read in Spain—tells us that the real risk is not that AI fails technically, but that it fails strategically: that it is implemented without transforming, adopted without integrating, and invested in without measuring.
That is the point where, if we are not critical, Mexico could get stuck: a country that boasts of its AI adoption but fails to convert it into a real advantage.
By Irene Marqués, Partner at the Olivia consultancy.