For more than a decade Colombia has been building one of the most ambitious open public procurement data systems in Latin America. SECOP II — the platform run by Colombia Compra Eficiente, the national procurement agency — now records more than a million procurement processes a year, with information on contracting entities, contract objects, awarded values, performance periods and supplier profiles. It is an unprecedented wealth of data for any developing country. And yet most of that information remains underused.
The problem is not the absence of data. It is the gap between the availability of information and the institutional — and technical — capacity to interpret it in time. The Office of the Comptroller General, the Inspector General’s Office and other oversight bodies work with limited teams against a contracting universe that grows every year. Civil society organisations carry out valuable oversight, but with insufficient resources to scale their analysis. And the media, which plays a fundamental role in accountability, operates reactively: it covers the scandal once it happens, not the warning sign that preceded it.
It is in that gap that artificial intelligence (AI) applied to public data offers its greatest value. Not as a replacement for human judgement or for oversight institutions, but as an amplifier of their analytical capacity: a system that can review all of an entity’s contracts at once, compare them against the sector average, detect statistically significant deviations and prioritise which ones deserve immediate review.
The potential of this technology in the Colombian context is concrete and measurable. Consider some of the best-documented risk patterns in the public procurement literature: the concentration of contracts in a single supplier, the repeated use of “manifest urgency” without proportionate justification, tender documents with technical requirements that artificially exclude competitors, or contracts whose value sits systematically just below the mandatory open-tender threshold. Each of these patterns leaves traces in SECOP II data that a trained algorithm can identify in seconds, whereas a human team would take days or weeks to detect them manually.
Advances in natural language processing (NLP) open an additional dimension. Tender documents — files of dozens or hundreds of pages that set the rules of each competitive process — can be analysed automatically to identify unusual clauses, requirements atypical for the sector standard, or restrictions that favour one specific supplier profile. What previously required a legal expert reviewing one process at a time can now be done systematically and comparatively across thousands of processes at once.
International experience supports this approach. In Mexico, the IMCO Contrata system has used data analysis to identify direct-award patterns in the health sector during the pandemic. In Chile, the Mercado Público portal — with characteristics similar to SECOP II — has been the subject of academic studies demonstrating the effectiveness of statistical analysis in predicting contracting irregularities. In the United Kingdom, the Cabinet Office has integrated risk-analysis tools into its internal audit of government contracts. The common thread is the same: the data already exists; technology makes it possible to use it intelligently.
For Colombia, this technological shift has implications that go beyond administrative efficiency. In a country where the Corruption Perceptions Index is deteriorating and where public distrust of institutions is at historic levels, demonstrating that the state can detect irregularities before they become a scandal carries considerable symbolic and political weight. Every early warning that prevents an irregular contract is, at the same time, a signal that the system works and that those who try to exploit it face real risk.
Technology alone, however, does not solve the structural challenges of Colombian public procurement. For data analysis to have real impact, at least three complementary conditions are required. First, quality and consistency in the information published on SECOP II: recording errors, inconsistencies in how contracting methods are classified, and late or incomplete publication of documents significantly reduce the predictive power of any analytical model. Second, institutional willingness to act on the alerts generated: an early-warning system that finds no counterpart willing to intervene is wasted technology. Third, opening data systems to external actors — civil society, academia, the media — so that analysis is not concentrated exclusively in government institutions.
The path towards more transparent, efficient and honest public procurement in Colombia necessarily runs through technology. But it also runs through the political decision to invest in analytical capacity, through strengthening open data systems, and through building oversight ecosystems where the state, civil society and the private sector collaborate instead of operating in silos.
LuxIA is a concrete bet in that direction. By cross-referencing SECOP II data in real time with risk models built on empirical evidence and technical-legal criteria, the platform allows oversight groups, suppliers and institutions to identify high-alert procurement processes, compare the behaviour of entities and sectors, and access analysis that previously required specialised teams and weeks of work. It is not the end of the road: it is the beginning of a new way of doing oversight in Colombian public procurement.