Five Minutes Behind in the AI 'Race' - Mexico Business News

STORY INLINE POST

Every time I talk to a business owner, the same phrase comes up, almost word for word: I feel like I'm way behind on AI. He sees others using GPT, automations, agents, all kinds of tools, and he feels like all he does is ask ChatGPT for simple things. That contrast creates something between guilt and anxiety, like there's an exam already underway and he hasn't even opened the notebook.

This isn't just an ego issue among fellow business owners. Micro-businesses make up 95.4% of Mexico's economic units and employ 41.4% of the workforce (National Institute of Statistics and Geography, INEGI, Economic Census 2024). When we talk about whether small and medium-sized businesses (PYMES) should adopt AI, we're really talking about the backbone of the Mexican economy. If the typical business falls behind not from incapacity but from bad information or misdirected fear, the cost isn't just that business' problem, it's a competitiveness cost for the whole country.

That's why I'm opening a four-part series with this article on AI implementation in traditional PYMES: restaurants, clinics, retail, small manufacturers, law firms, construction companies. Not to sell you the fantasy of an agent that handles service, quotes, and closes sales while you sleep, but to ground where the real value is, where the narrative is oversold, which objections are legitimate and which are excuses, and what a well-made decision looks like when you don't have a technical team or unlimited budget. This first article is the simplest, and I think the most necessary: clarifying what "falling behind" actually means before we get into the rest.

I understand why my clients feel this way. But the comparison is flawed from the start. GPT is just one link in the whole universe of artificial intelligence, not the entire universe. In fact, of all web traffic toward AI solutions in Mexico, 77% goes to consumer generative AI (like ChatGPT), 20% to models or application programming interfaces, and only 2% to open-source platforms or in-house development (Economic Commission for Latin America and the Caribbean, ECLAC, and the National Center for Artificial Intelligence, CENIA, Latin American Artificial Intelligence Index, 2025). In other words, almost the entire country is using the most superficial layer of AI. The deep changes aren't in using a large language model (LLM), they're in what you build around it. And that difference, subtle as it sounds, is what separates someone who uses AI from someone who builds with AI.

Using AI Isn't the Same as Building With AI

At TAND, we did it step by step, and that's the proof of what I mean. We started like everyone else: using language models for isolated tasks. Today, we have a full suite that brings together user experience, content creation, patient tracking, care protocols, and proprietary systems development. The LLM is still there, but it's no longer the main character, it's one piece within something bigger. The result isn't a nice anecdote: we now serve 30% more patients with the same staff, and with better reviews. Revenue per hour has also gone up 25%.

That didn't happen from using GPT better. It happened from ordering, first of all, the most boring parts of the business: a response catalog, historical data, tables that had been disorganized for years, and training the models on real business information. Nobody creates content about that because it isn't sexy. But it's exactly what determines whether AI adds value or just adds noise on top of the chaos you already had.

Being Ahead Doesn't Mean What You think

I worked with a bank that had major issues with customer service and user experience, its systems weren't the best, and its app and digital services lacked good experience design. And yet, its customers were happy. It was behind on systems, ahead on what actually moved the needle with its people.

The opposite case also exists, and it's more dangerous. I worked with an architecture firm that handled dozens of projects a year, with robust systems like Autodesk Construction Cloud (ACC) and broadly integrated management tools. But its processes were inefficient and, at times, redundant. Putting AI on top of that wouldn't have fixed anything, it would have accelerated the same redundancies and inefficiencies it already had, just faster and with more appearance of sophistication.

And before you assume your business is at a disadvantage compared to the rest of the country, it's worth looking at the real baseline most businesses operate from. According to the National Business Financing Survey (ENAFIN 2024, INEGI), only 54.3% of Mexican businesses accepted card payments in 2023, and the rest didn't, mainly because they preferred cash, customers didn't ask for it, or the cost or transaction amounts didn't justify it. Assuming that same business is going to prioritize predictive analytics or autonomous AI agents before solving something as basic as how it takes payment is confusing what the technology vendor wants with what the owner's reality actually is.

Yes, It's a Race, Just Not the Way You Think It Is

This is where people are surprised. When I tell them there's no such race, I'm actually understating it. It is a race. It has already started. But the lead held by whoever started earlier isn't as big as it looks, we're five minutes behind, not five years. And with a clear strategy and a disciplined pace of implementation, you don't just catch up to the ones making noise, you pass them by a wide margin.

And there are real numbers behind this, not just my opinion. Mexico is classified as an "adopter" in the Latin American Artificial Intelligence Index (ILIA 2025, ECLAC and CENIA), scoring 47.03 points and ranking eighth regionally: above the regional average, strong in research and development, but still at zero on vision and institutional framework because the country doesn't even have a formal national AI strategy yet.

Here's the real trap: In the media, it's all AI, AI, AI, and that creates anxiety and FOMO (fear of missing out). But once you look at the detail, the story changes. The Massachusetts Institute of Technology (MIT), through its Media Lab and Project NANDA, published a 2025 report titled, "The GenAI Divide: State of AI in Business," based on an analysis of more than 300 public deployments, 52 organizational interviews, and 150 executive surveys. The conclusion: 95% of generative AI pilots at companies fail to deliver measurable financial impact. And Gartner, in June 2025, was even more specific about what's coming: it predicts that more than 40% of agentic AI projects will be canceled by the end of 2027, not because of the model's capability, but because of spiraling costs, unclear business value, and inadequate risk controls.

So yes, almost everyone is using AI in some form. But actually getting good at using it, that's the real bottleneck, and that's exactly where most fall short. The contention that "everyone already has this mastered" is more noise than reality.

I don't have this fully figured out, and I don't claim to. What's hardest for me, even now, is the speed at which everything changes. Today it's Codex, tomorrow it's Claude, then Kimi, then Nvidia releases something new, and on it goes. Any specific answer about which tool to use expires in weeks.

That's why I believe the real risk isn't failing to bring AI into your business. The risk is that you, as the owner, don't understand the scope of what AI can do in your company and don't have a clear strategy for where it actually makes sense to use it, because spoiler, it's not everywhere.

In the next article of this series, I'll dig into that overselling directly: where the AI-for-PYMES narrative is inflated and where there's genuinely real time or money saved, with data and no hype.

For now, the question isn't whether you're behind, but rather, where do you think you are today, and how do you figure out what your next step actually is?