AI nobody measures pays off for nobody
Almost half of Argentine SMBs already use AI, but most can't tell if it works. Before adding another tool, pick one process and measure it.
The first large survey on AI in Argentine SMBs is out. It was done by nadIA, an initiative of UTDT and Fundar, with support from the IDB and Observatorio PyME. They surveyed 402 companies, mostly manufacturing and software, in November and December 2025.
The headline is good: AI is no longer for a few. What comes after is what I care about.
The numbers
- 41.6% of SMBs already use at least one AI tool.
- Among those, 57% don't formally measure whether it pays off.
- Only 4% of all firms have a dedicated AI budget, and just 3.3% have a written policy on how to use it.
- 56% of adopters are in an experimental stage: they try things, but without the skills or rules to sustain them.
- Asked about the next five years, 77% expect higher productivity. Today, only 16.8% say they've seen it.
What I see in practice
I see this a lot. A company adds a chatbot, someone uses ChatGPT for emails, someone else writes summaries, and six months later nobody can say which task got faster or how many hours were saved. Not because it didn't help, but because nobody measured the before.
The gap between what's expected and what's seen isn't a model problem. It's about how AI gets into the company: no chosen process, no starting number, and nobody checking if things improved.
What I suggest
Before adding another tool:
- Pick one process. Something that repeats every week: entering invoices, answering the same questions, building the monthly report.
- Measure how long it takes today. Hours per week and how many people touch it. A spreadsheet is enough.
- Automate it with AI and measure again. If it dropped, scale it. If not, cut it and try another.
You don't need a big-company budget or an AI team. You need to know what you want to improve and have the before number.
My bet
The SMBs that get real value from AI won't be the ones that try the most tools. They'll be the ones that measure one task, improve it, and repeat.
Full report (in Spanish): AI adoption survey in Argentine SMBs (nadIA, PDF).
More pieces
Does this sound like a process of yours?
I write this from real projects. If something here resonates, tell me about it on a 30-minute call.