AI in Argentina: fintech, agritech, and the Buenos Aires engineering hub pulling the country's most opinionated AI buyers off legacy stacks.

Argentina does not yet have a binding AI-specific statute, but the regulatory floor is thicker than it looks. The AAIP's Resolución 161/2023 sets recommendations on trustworthy AI that are treated as the operational standard in enforcement actions under Ley 25.326. Disposición AAIP 2/2023 raises the bar for biometrics and high-risk processing — that is the binding instrument for face-recognition, automated hiring decisions, and credit scoring. The BCRA Comunicación A 7724 governs automated channels in financial services. A Proyecto de Ley de Inteligencia Artificial has been introduced in the Cámara de Diputados and would establish a risk-tiered framework loosely modeled on the EU AI Act if enacted, though the legislative timetable through 2026 is uncertain. Companies should budget compliance work as if a binding statute lands within twelve to eighteen months — the direction of travel is clear even if the exact text is not.

Three factors compound. First, the public universities (UBA, UNLP, UTN, UNC) graduate large cohorts of mathematicians, statisticians, and software engineers, and the country has historically punched above its weight in math olympiads and ML research. Second, the macroeconomic environment has pushed dollar-billing services work — outsourcing, nearshore engineering, AI consultancy — to be one of the few sustainable career paths for senior technical talent, which means the supply of mid-to-senior engineers willing to work on AI projects is unusually deep relative to GDP. Third, Globant scaled a local-but-global AI services model from Buenos Aires that has trained a generation of project managers, ML engineers, and AI product leads who now run boutique consultancies or in-house teams across Argentina, Uruguay, and Chile. The practical implication for a buyer: you can staff an Argentine AI team faster and at higher senior density than in nearly any other LATAM market.

The resolution operationalizes what Article 1 and Article 27 of Ley 25.326 require for automated decision-making. In practical terms, it asks for three things from any AI deployment that materially shapes a consumer outcome — credit, insurance, employment, public-facing recommendations. First, transparency: the company has to be able to explain, in plain Spanish, the logic involved, the data inputs, and the foreseeable consequences of the decision. Second, human review: where a decision has legal or similarly significant effects, the consumer has the right to request human intervention and to contest the outcome. Third, documented impact assessment: while framed as a recommendation, the AAIP increasingly treats the absence of an Evaluación de Impacto en Protección de Datos as a sanctionable gap in due diligence. The cost of compliance is small. The cost of an AAIP inspection after a consumer complaint is not.

A mid-market AI project — 200 to 2,000 employees, headquartered locally — typically runs six to nine months end-to-end to production, broken into roughly six weeks of discovery and data-readiness, eight to twelve weeks of build, four to six weeks of pilot with a controlled user cohort, and the remainder for full rollout and stabilization. Argentina-specific friction tends to concentrate in three areas: (a) the AAIP impact-assessment cycle for any high-risk use, which adds two to four weeks if the legal team is new to the topic; (b) integration with locally-rooted ERPs like Tango or Bejerman, which often lack modern APIs and require middleware; and (c) the Spanish-language quality bar in Río de la Plata dialect, which usually requires a domain-specific evaluation set built from real customer interactions rather than off-the-shelf benchmarks. Where the work moves faster than Brazil is on data infrastructure — Argentine mid-market companies are typically further along in ERP and data-warehouse modernization than their Brazilian counterparts.

Most serious Argentine AI vendors bill in US dollars through offshore structures (typically Uruguayan or US LLCs) to insulate clients from the local FX regime and inflation indexation. If you contract a vendor invoicing in Argentine pesos under MEP or CCL exchange rates, you take on currency risk that compounds quickly on a multi-month engagement. Three practical defaults: (a) contract in USD with the offshore entity, with a clearly defined services-from-Argentina clause for tax treatment; (b) pay milestones rather than monthly time-and-materials so currency movement doesn't silently re-scope the project; (c) include a data-residency clause specifying where customer data is processed — vendors using US-region AWS or Azure are the safest default, while vendors hosting in Argentina need to demonstrate they meet your home jurisdiction's data-transfer requirements. None of this is unique to AI work, but the long tail of Argentine vendor relationships rewards attention to these terms upfront.