The Superintendencia Financiera has integrated artificial intelligence into its anti-money laundering supervision system. The algorithm cannot process what was never digitized. And in the vast majority of supervised entities, the SARLAFT onboarding form still arrives handwritten from the field.
What the Superintendency announced
The SFC activated an AI tool to strengthen the detection of suspicious transactions related to money laundering and terrorism financing (LA/FT). The announcement is part of the entity's Digital Supervision Strategic Plan, aimed at more proactive and scalable oversight of the Colombian financial system.
The move is consistent with the global trend. According to McKinsey, banks that integrate AI into their AML systems reduce between 20 and 30 percent of false positives in suspicious transaction detection models. Supervisory efficiency improves. Regulatory risk drops.
The blind spot the announcement does not resolve
The SFC's AI operates on structured data: transactions, traces, digital behavior. But the SARLAFT onboarding process at fiduciaries and supervised entities does not begin with structured data. It begins with a paper form, signed by hand, filled out in the field by a compliance officer or the beneficiary themselves.
That form contains the client's signature, the declaration of funds origin, the legal representative's information, a fingerprint in some cases. It is the source document for the entire compliance file. And in most supervised entities, that document reaches the compliance team as a scanned image, when it arrives at all.
No AI system can issue a risk verdict on a document it cannot read.
The SFC's algorithm detects patterns in transactions. The handwritten field form is the document that comes before the system. That gap is the bottleneck of SARLAFT compliance.
The cost of that blind spot at the next SFC visit
Asobancaria has documented that AML compliance costs in the Colombian financial sector have grown steadily, driven by tightening documentary evidence requirements. The regulator does not only want systems to detect: it wants entities to demonstrate that the client file is traceable end to end.
A fiduciary that can show a digitized onboarding form, validated field by field, with an automated compliance verdict, is in a very different position before the supervisor than one that presents a scanned image with handwritten margin notes.
From field form to auditable decision
DocIntel takes the handwritten field form, applies OCR with calligraphic profiling, validates each field against regulatory rules (name, ID, funds origin, restricted lists), and issues a structured verdict: the file is complete, there is an inconsistency alert, or the case requires manual review.
The result is not a digitized file. It is an auditable decision on that document. The client consumes the verdict, not the software that generated it.
For SARLAFT compliance teams, that means moving from reviewing scanned folders to receiving structured risk-level alerts, with the complete file ready for the SFC visit.
The moment is now
The SFC just raised the supervision standard with its own AI. Supervised entities that do not resolve the field document gap in the next twelve months will arrive at the next visit with a visible asymmetry: the regulator has AI, the client file is still paper.
Forty-five minutes is enough to watch DocIntel whether this gap exists in your process and what it takes to close it.