TechMex Talks, Episode 3: "Technology-using companies and AI"
When an organization says it wants to implement artificial intelligence, it is rarely declaring an objective. It is declaring an anxiety.
That was the underlying tension in the third and final episode of TechMex Talks, "Incorporating Artificial Intelligence into the Technology Ecosystem", organized by TechMex Austin, the Austin chapter of Red Global de Mexicanos y Mexicanas en el Exterior, in collaboration with Software Guru. The conversation, held on March 17, 2026 at Improving's offices in Austin, brought together Eva Noyola, Ph.D, Data Scientist at the Texas Comptroller of Public Accounts, and Javier Carro, CTO of BluePeople, moderated by Gerardo Pelayo, Ph.D, Chief Research Officer at Service Council.
If the first two episodes explored how AI transforms the technology professional and reconfigures the entrepreneur, this third panel addressed the question that closes the loop: what happens when artificial intelligence meets the operational reality of the organizations that consume it?
Perception is not adoption: the 50% mirage
Gerardo Pelayo opened with a data point that framed the entire conversation. In an annual Service Council survey of more than 3,000 field service engineers, 50% believe AI could be very useful in their work environment, but it is not useful yet.
This is not about resistance. Pelayo was emphatic: "When tools make sense, they are welcomed. When they are imposed, when they are not aligned with the incentives of the people executing in front of the customer, that is when resistance problems appear."
Eva Noyola confirmed the pattern from the public sector: "There are enthusiastic users, but the tools have not reached their place yet. So we are seeing shadow IT. People are so eager to use these tools that they will go use them somewhere else, which can become a serious security problem."
AI is not a strategy. It is part of the strategy.
Pelayo stated the point directly: "I am one of those people who believe artificial intelligence cannot be a strategy. It has to be part of the strategy."
Javier Carro brought the idea into practical terrain: "We have been offered AI marketing tools that promise to reach one million people with personalized emails. But in our industry, if I send ten emails and get zero replies, then even if I send one million, they will still be zero."
Any factor multiplied by zero is still zero. If the underlying process does not work, AI only accelerates inefficiency. Or as Carro summarized it: "It is very easy to automate garbage. If you are doing something badly, then you will do it badly faster."
The public sector as a mirror of truth
Noyola shared a study that evaluated AI adoption in the public sector across ten countries. The results were counterintuitive: "The ones that are most advanced are Saudi Arabia, South Africa, India, Brazil. The conservative ones are France, Germany, Japan, and us."
Why? "The laws are much stricter in these countries. The risks of moving too fast and letting something slip are much more costly."
The tension becomes personal: "Technically I have all public dollars at my fingertips. Can you imagine? Not even drunk would I put half a data point from my tables into any AI model."
The discourse changes faster than implementation
Noyola described how the dominant AI discourse has changed radically:
"The first year was: AI is going to fix your life. The second year: forget that, agents, agents are going to solve your life. The third year, surprise: we have a human layer in our AI models, we are not going to let them run by themselves."
The problem is not only that the technology evolves quickly. It is that the commercial discourse around the technology evolves even faster, creating expectations that even the industry itself cannot sustain.
User experience: the data no one wants to see
Pelayo shared an alarming trend: even though efficiency has improved, the share of technicians who say their day-to-day work has become easier dropped from 73% to 68% to 61% over three consecutive years.
AI can improve company metrics while worsening the experience of the people who use it. KPIs look good on the executive dashboard; the field technician feels that their work has become harder.
Trust: the infrastructure you cannot buy
Pelayo articulated it with precision: "Trusting the data is not the same as trusting the intelligence that comes out of the data."
Carro reinforced the point: "Companies implement something and then roll it back because it did not work for them. You have to be very careful about putting the right safety rails in place so that, for certain critical decisions, a person gives the final approval."
The human in the loop is not a concession to nostalgia. It is trust architecture.
Incremental improvement vs. transformation
Pelayo offered one of the clearest distinctions: "Where I separate incremental improvements is this: I am working on the same question, but with better answers. The moment I am answering new questions, I am doing transformation."
Through this lens, most AI implementations sit in the incremental improvement category. Calling it transformation inflates expectations and dilutes the meaning of the term.
Closing the loop
Taken together, the three episodes draw a sequence that is also a warning: the professional who does not adapt loses relevance; the entrepreneur who does not integrate AI loses competitiveness; the company that implements AI without understanding its own users loses trust.
The full series leaves an architecture for decision-making: first individual judgment, then business strategy, and finally the organizational capacity to absorb change. Without that sequence, AI becomes what Pelayo warned about at the start: a solution looking for a problem.
Final reflection
Are we designing AI to serve organizations, or are we redesigning organizations to serve AI?
The answer should not require reflection. But the fact that it does is precisely the diagnosis.
The most powerful AI in the world is useless if the person who needs to use it does not trust it, does not understand what it does, or does not have room to say it is not working.
That is the question left open. Not for a fourth episode, but for every organization deciding its next step.
This article is part of the TechMex Talks series, "Incorporating Artificial Intelligence (AI) into the Technology Ecosystem", an initiative by TechMex Austin in collaboration with Software Guru.
This content was created with support from Artificial Intelligence tools and human editorial curation, in alignment with the principles discussed during the panel.