Agentic AI in Latin America: 84% of companies already rank it among their top three priorities
84% of companies in Latin America already rank agentic AI among their top three priorities. It is the AI that uses agents: systems that plan and carry out tasks with some autonomy. But prioritizing is not scaling: most still see moderate gains, and among those that do not exceed a 5% return, the main barriers are about management, not technology.
That is the conclusion of the study “Artificial intelligence in Latin America, on the way to a new scale of value and competitiveness”, by MIT Technology Review in Spanish and NTT DATA. It is based on a survey of 300 business leaders in Argentina, Brazil, Chile, Colombia, Mexico and Peru and on ten interviews (MIT Technology Review in Spanish, 2026, in Spanish; NTT DATA, 2026, in Spanish).
According to the study, companies in the region are leaving the testing stage behind to scale AI (MIT Technology Review in Spanish, 2026, in Spanish).
What the data says
| Indicator | Figure | Source |
|---|---|---|
| Companies that rank agentic AI among their top three priorities | 84% | MIT Technology Review in Spanish, 2026 |
| Companies at the “mature” or “advanced” levels of traditional AI | 54% (35% a year earlier) | MIT Technology Review in Spanish, 2026 |
| Companies at the “mature” / “advanced” level of generative AI | 31% / 16% in 2025 (9% / 5% in 2024) | MIT Technology Review in Spanish, 2026 |
| Companies at the “mature” or “advanced” levels of agentic AI | 44% | MIT Technology Review in Spanish, 2026 |
| Companies with a well-defined overall AI strategy | 66% | MIT Technology Review in Spanish, 2026 |
| Companies with a specific generative AI / agentic AI strategy | 42% / 37% | MIT Technology Review in Spanish, 2026 |
| Companies making significant investments in AI | 79% (33% describe them as very significant) | MIT Technology Review in Spanish, 2026 |
| Companies with at least a 5% increase in profits attributable to AI in the last fiscal year | 80% | MIT Technology Review in Spanish, 2026 |
Source: MIT Technology Review in Spanish, 2026, in Spanish.
Two caveats for reading these figures correctly:
- They are self-assessments. Companies rate their own maturity level; the article gives no earlier figure for agentic AI to compare against.
- NTT DATA is co-author and sponsor. It produced the study with MIT Technology Review in Spanish, and the article presenting it is sponsored content. The survey dates are not reported either.
AI gains for companies in the region: real, but moderate
54% of organizations report increases of 5% to 15% in profits attributable to AI, 22% between 16% and 40%, and only 4% exceed 40%. The most common gains are in productivity, operational efficiency and automation (MIT Technology Review in Spanish, 2026, in Spanish).
Interactive chart
| Increase in profits | Organizations |
|---|---|
| Less than 5% | |
| Between 5% and 15% | |
| Between 16% and 40% | |
| More than 40% |
What holds back the adoption of agentic AI
Among organizations that do not yet achieve returns above 5%, the main barriers are not technological (MIT Technology Review in Spanish, 2026, in Spanish):
- Change and project management: 36%.
- Resistance within the organization: 34%.
- Integration with existing systems: 33%.
In addition, 44% say that training their employees is their main challenge in accelerating adoption. And while 85% consider their investment in applications sufficient, 31% rate their investment in technology infrastructure as insufficient, more than twice the share for applications (12%) (MIT Technology Review in Spanish, 2026, in Spanish).
What makes AI pay off
For 42%, the main driver of return is data quality and governance models, followed by the ability to put AI into production and scale it (40%). And although 87% consider trust and security relevant or very relevant to the success of AI, fewer than half say they are fully confident in the strength of their governance, their cybersecurity or their ability to meet data sovereignty requirements (MIT Technology Review in Spanish, 2026, in Spanish).
Prioritizing is not scaling
Another NTT DATA study, this time with CIONET, helps put the 84% in perspective. The Agentic AI Observatory in Spain and Ibero-America 2026, published in June and based on interviews with 130 organizations, reports these figures (NTT DATA and CIONET, 2026, in Spanish):
- Only 3.8% have taken agentic AI to an industrial-scale implementation; 87.7% are still experimenting and 59.2% are running scattered pilots.
- 80.8% have no operational mechanisms to control the autonomy of their agents, and only 6.9% have an advanced governance framework.
- 40% say their main challenge is building business cases with measurable economic impact.
- 67% have trained less than 30% of their staff in AI.
The samples are different and the Observatory’s includes Spain, so the figures cannot be compared one to one. Even so, the message is the same: agentic AI is already on almost every company’s agenda, but very few run it at scale and with control.
What it means for companies in Ecuador
Ecuador is not part of the MIT Technology Review and NTT DATA study, and the Observatory does not specify the countries it covers. Even so, Ecuadorian companies compete with those in Colombia, Peru and Mexico, which are included. If the vast majority of companies in the region already prioritize agents, the advantage will not lie in having them but in making them work well. The study shows where that advantage is decided: organized data, systems that can be connected and prepared teams. These are also the fronts emphasized by ILIA 2026 (training, data and collaboration), an index in which Ecuador ranked ninth out of 20 countries (Wasivia, 2026, in Spanish).
How to prepare your company for agentic AI
- Review your data before your tools. An agent performs only as well as the information it can access.
- Take an inventory of your systems. Identify which ones can be connected to an agent and which need prior integration.
- Plan the change, not just the project. Explain to the team what the agent will do, what each person will keep doing and how the result will be measured.
- Invest in training. It is the main challenge for almost half of the companies in the region and, according to the Observatory, two in three organizations have trained less than 30% of their staff (NTT DATA and CIONET, 2026, in Spanish; Wasivia, 2026).
- Start with a narrow, measurable agent before expanding to other areas, as the companies that already scale their agents do (Wasivia, 2026; Wasivia, 2026).
At Wasivia, we help define the AI roadmap and prepare data and systems through our AI consulting and strategy, and we train teams through corporate AI training.
References
- MIT Technology Review in Spanish — 84% of Latin American companies already prioritize agentic AI (Nov 16, 2026, sponsored content by NTT DATA), in Spanish
- NTT DATA — AI in Latin America 2026 (study page), in Spanish
- NTT DATA and CIONET — Agentic AI: only 3.8% of companies in Spain and Ibero-America have managed to scale it (Jun 16, 2026), in Spanish
- Wasivia — AI agents in production: 79% of companies already use them, but only 20% make full use of them
- Wasivia — Ecuador in ILIA 2026: 51.8 points and ninth place in AI in the region, in Spanish
- Wasivia — AI training: why company culture matters more than the tool
- Wasivia — What are AI agents and why are they transforming businesses?
