Microsoft is trying to move artificial intelligence in Mexico from individual desktops to the decisions companies make, Expansión reported on October 29 after the Microsoft Frontier Tour at Expo Santa Fe in Mexico City. Ivonne Maritza Mejía, general director of Microsoft México, told the audience that people are ready for AI and that the remaining challenge is organizational. Jessica Hawk, a corporate vice president of product marketing for Azure, laid out the company’s approach, and a demonstration showed software agents preparing a supplier decision that a manager approved before any system was updated. The demonstration used a fictional company, and Expansión noted that it offered no measured savings, error reduction or return on investment from a real client. Lens found no second report of the event, so the details below rest on Expansión.

The demonstration

The demo followed a fictional company called Caldova. A supplier manager used Copilot in Excel to query Fabric, Microsoft’s data platform, and review 11 supplier candidates, including their financial risk. According to Expansión, the system matched records against selection criteria found in emails and meeting transcripts, showing which suppliers qualified and why others were rejected.

The manager then asked the system to evaluate the proposals, prepare materials and organize a meeting. Before the system updated Dynamics, Microsoft’s enterprise resource planning software, the manager reviewed and approved the evaluation. Expansión described that approval as a human checkpoint, and it is the feature that connects the demo to the supervision themes that ran through the rest of the event.

The pitch: Copilot, IQ and Agent 365

Hawk said the work of adopting AI starts with defining the outcome an organization wants, then reviewing which teams take part, which decisions they make and where they get stuck. She said implementation should involve operations leaders as well as the technology department, according to Expansión.

Models trained on public data do not know a company’s agreements, clients or decision criteria, Hawk said. Microsoft’s answer is a platform called IQ, which brings an organization’s own knowledge to Copilot, applications and AI agents. Expansión described the portfolio as three parts: Copilot for employee interaction, IQ for business context and Agent 365 for managing and supervising agents.

The supervision piece carried much of the argument. Organizations need to know which agents are working, who is accountable for them, what they cost and what value they add, Hawk said. Technical teams, she added, must be able to trace the decisions agents make and intervene when an agent acts outside expectations.

The adoption numbers behind the push

Expansión cited Microsoft’s AI Diffusion Report for the share of people aged 15 to 64 in Mexico who used generative AI tools in the second quarter of 2026: 21.4%, against 18.8% globally. The report’s earlier edition put Mexico at 20.1% for the first quarter and the global rate at 17.8%, with Mexico ranked 56th among 147 economies, according to Mexico Business News on June 19. Microsoft’s method relies on aggregated, anonymized telemetry adjusted for population, internet penetration and device-market differences, Mexico Business News reported. These are modeled estimates from the vendor, not a government survey.

At the June briefing, a Microsoft México public affairs executive named Pliego (Miguel in Mexico Business News, Manuel in El Cronista) said the country still has to grow its adoption and needs more talent training, connectivity, energy and digital infrastructure and regulatory frameworks, according to both outlets.

From Microsoft’s Work Trend Index 2026, Expansión reported that 67% of Mexican AI users say they do work they could not do a year ago, while 28% see clear alignment from their leaders on changing how the company operates. Intelligent CIO reported the same two figures on September 21. The same survey, according to that outlet, found that Microsoft attributes 67% of AI’s impact to organizational factors and 32% to individual effort. Ezequiel Glinsky, Microsoft’s director of technology for Latin America, said the next step is to redesign processes and ways of collaborating so that progress produces results at scale.

Lens analysis: the 67% and 28% figures describe the gap Mejía and Hawk spent the event on. Users say they are doing more, but few say their leaders have lined up the organization behind it. Both numbers come from Microsoft’s own surveys, and Lens did not open the underlying reports.

Where the data lives

Data location was one of the topics Hawk addressed. She said the right choice depends on the needs of each operation, and that keeping data inside the country is only one part of controlling models, infrastructure and administration, according to Expansión. She described public sovereign, private and local-partner cloud options. Expansión noted that Microsoft has operated a cloud region in Querétaro since 2024.

That region came with a larger commitment. Bloomberg Línea reported on September 24, 2024 that CEO Satya Nadella announced $1.3 billion of investment in Mexican AI infrastructure over three years, along with a goal of training five million people in AI skills in the same period. Bloomberg Línea said Microsoft had opened its first Azure data center region in the country, in Querétaro, the previous May. Expansión’s 2024 coverage carried the same $1.3 billion and five million figures.

The infrastructure and rules around it

Agents that run on cloud capacity need power and permits. Mexico Business News reported on June 23 that the Mexican Data Center Association, MEXDC, expects $82.5 billion to go into data center construction and equipment from 2026 to 2031. Mexico had 279 megawatts of installed capacity, 205 megawatts under construction and 1,730 megawatts announced, the association said, and Querétaro held 72% of existing capacity.

The same article quoted MEXDC executive director Adriana Rivera saying the system needs about 1.7 gigawatts of additional energy capacity, and it reported that MEXDC estimates about five years pass between permitting and operation. The article also said more than 60% of the transmission network runs near maximum capacity, with bottlenecks in the Bajío, Nuevo León and the northern border states, and that installation costs in Mexico are about four times those in Brazil.

Regulation lags behind adoption, according to a separate opinion piece. Aldo Ricardo Rodriguez Cortes, CEO of the firm Lawgic, who leads the Observatorio IA México, wrote in Mexico Business News on October 15 that 130 federal AI initiatives have been registered since April 2023, that fewer than 10 have been approved in either chamber and that none of the comprehensive bills allocates funding. The most advanced bill has 223 articles and would create a national AI authority. He noted that Congress lacks express constitutional authority to pass a general AI law and that the Senate wants one passed by the end of 2026.

His conclusion was that “Mexico does not have a regulation gap. It has an implementation gap.” That is the author’s view, not a finding by a regulator. The enacted measures he lists are narrow: a 2025 overhaul of the industrial property law and a May 2026 reform on labor and copyright that protects voice and image from generative AI.

What to watch

Lens analysis: the event gave Mexican companies a vocabulary of outcomes, decisions, checkpoints and agent accountability, but no evidence yet of results. Expansión flagged the missing numbers itself. A useful signal would be Microsoft naming Mexican customers and publishing measured outcomes for agent-based workflows, which the demo did not.

Adoption data is the second marker. Microsoft’s next AI Diffusion report will show whether Mexico’s 21.4% keeps climbing relative to the global rate. Third is the legislative calendar: the Senate’s goal of passing an AI law by the end of 2026 depends on a constitutional amendment that, according to the Observatorio IA México, has been proposed 16 times since 2023 without approval. Finally, the data center buildout depends on grid capacity and permitting times that MEXDC itself describes as constraints.

Sources

Image: Santa Fe Skyline 2017 CDMX.jpg by Gobierno CDMX * https://www.flickr.com/photos/140660272@N07/, CC0, via Wikimedia Commons. Illustrative photo; it does not show the events described.