AI in healthcare isn't bought, it's measured: Colombia has the data but has yet to prove it frees up clinical time

By Portafolio · October 31, 2026.

Colombia surpassed 7.7 million Digital Care Summaries (RDA) received in June 2026, according to a Ministry of Health report cited by Portafolio. An RDA is an extract of a patient's medical record that can be shared between different institutions. A total of 744 healthcare providers, with 2,456 sites, were taking part, and the records covered some 2.7 million patients.

The figure comes alongside the report "Intelligent Transformation of the Healthcare Ecosystem 2026–2030" from OBS Business School. According to the document, AI can reduce the administrative burden on doctors and nurses, support clinical decisions, improve patient follow-up and allocate resources more judiciously. But it makes this conditional on three requirements: reliable data, staff training and human oversight.

The article points to specific risks. One is buying tools without defining what need they address or without testing them across different departments. Another is automating processes that already work poorly. And a technical test only shows that the system works, not that it improves clinical practice, shortens wait times or benefits the patient.

That is why OBS recommends setting indicators and a baseline before rolling out a solution. Proposed metrics include safety, care times, patient experience and workload. Validation should be carried out under real-world conditions, with professionals and data from the setting where the tool will be used, and integrated into standard workflows so as not to create parallel circuits or duplicate records.

There is also a question of accountability. Juan Ramón Segura (Medtronic) stresses that sharing information between hospitals and other stakeholders allows for better coordination of care. J. Alfonso Formáriz (Havas Health Network Spain) insists that human oversight matters more when a decision requires clinical judgment or a conversation with the patient. He also proposes governance that combines clinical, technological, legal, ethical and management profiles, with patients where appropriate, and with resources to monitor and update the systems.

**Our take.** The article offers hardly any impact figures: it doesn't say how many hours are freed up or how many hospitals use AI today. The 7.7 million RDAs measure interoperability, not clinical benefit. Confusing the two would mean falling into the very error the report describes: taking for granted something that has merely been installed.

Our thesis is that the bottleneck for healthcare AI is not the model but organizational trust: clean data, prior metrics and someone with the authority to shut the system down. It's the same pattern we see with enterprise AI agents, where the permissions and accountability layer matters more than raw capability.

Overall, the short term will be uneven. Some hospitals will buy tools because they are fashionable and see no improvement, and a gap will open between large and small institutions. Those with interoperable data and measurement discipline will move ahead first.

Even so, the path ahead warrants optimism. If AI absorbs the paperwork, giving minutes back to patient contact is a real first step. In the longer term, that same shared data infrastructure is what will enable earlier diagnoses and more preventive medicine, which is where the promise of eradicating diseases lies. Colombia has laid the foundations. What remains is to prove, with numbers, that clinical time is being freed up on top of them.