Jun 22, 2026

Mapeo de herramientas (IA) en contextos laborales: El caso de los Call Centers en Colombia

Daniel Bravo

La incorporación de la inteligencia artificial (IA) en entornos laborales ha reconfigurado las relaciones de trabajo. Pese a su magnitud y a ser pionera en tecnologías de monitoreo, la industria de centros de contacto (call centers) en Colombia permanece poco explorada. Este artículo realiza un mapeo exploratorio, mediante web scraping, de los principales actores y herramientas de IA del sector, con un inventario de 63 herramientas desplegadas por 12 operadores. Los hallazgos evidencian una marcada asimetría de información a favor de las multinacionales, la presencia de sesgos de autorreporte corporativo, así como la predominancia de herramientas orientadas a la gestión algorítmica. El despliegue actual sugiere complementariedad, con una tendencia reciente hacía el desplazamiento conducido por la automatización en la atención al cliente, aunque el presente inventario no permite inferir efectos netos sobre el empleo.

Review Project

See Code

View Related Sprint

Reviewer's Comments

Este proyecto aborda un problema real y relevante: la incorporación de herramientas de inteligencia artificial en el sector de call centers en Colombia, en un contexto laboral donde pueden surgir riesgos de vigilancia, gestión algorítmica y pérdida de capacidad de negociación de los trabajadores. Su principal aporte está en mirar un sector poco explorado en el país y construir un primer inventario abierto de herramientas, actores y relaciones entre proveedores y operadores. La metodología es trazable y útil como punto de partida, y el autor es claro y honesto al presentar su resultado como un mapa exploratorio, no como una medición completa del fenómeno.

Vale la pena situar el aporte frente a lo que ya existe. La vigilancia algorítmica laboral, como problema, ha sido ampliamente documentada, incluso en Colombia, por ejemplo, el proyecto Fairwork ha auditado las condiciones del trabajo mediado por plataformas en el país. Lo que este proyecto agrega de propio no es abrir un tema nuevo, sino llevar esa mirada a un sector aún poco explorado a nivel de inventario: los call centers. Citar de forma más explícita esos trabajos vecinos ayudaría a ubicar mejor su contribución.

En cuanto a las fuentes, buena parte de la información proviene de reportes corporativos y materiales públicos de las propias empresas, lo que tiende a subrepresentar las herramientas más sensibles: monitoreo directo, control del trabajador o vigilancia remota. El propio autor reconoce esta limitación e identifica las fuentes que quedaron por fuera, por ejemplo denuncias, información sindical, datos gubernamentales y registros judiciales. Por eso, más que señalar un vacío que el trabajo ya advierte, vale la pena subrayar que incorporar esas fuentes independientes en una versión futura sería clave para equilibrar el sesgo de autorreporte.

Desde la perspectiva de gobernanza y rendición de cuentas, el siguiente paso sería conectar el inventario con una pregunta de auditoría: qué herramientas tienen mayor potencial de afectar derechos laborales, quién las supervisa, qué garantías existen para los trabajadores y qué evidencia queda cuando una decisión laboral se apoya en estos sistemas. El proyecto tiene valor como base inicial, pero ganaría fuerza si pasara del mapeo descriptivo a una clasificación de riesgos y a criterios mínimos para auditar el uso de IA en entornos laborales.

This is a thoughtful and well-motivated piece of work that addresses a sector often left out of mainstream AI safety conversations. The topic is well chosen, the research instinct is sound, and the open dataset represents a genuine contribution that others in the region can build upon. The literature grounding is appropriate, and there is a clear sense throughout that the author understands the broader stakes of the problem being explored. Where the work could be further refined is in ensuring that the interpretive conclusions rest as carefully on the evidence as the more descriptive observations do, since some readings reach slightly further than the data, as currently presented, is positioned to fully sustain.

On the methodological side, the design is coherent and the author shows a commendable degree of self-awareness about the limitations of the sources used. A future iteration might benefit from adding a brief note on how collected entries were cross-checked, which would give readers greater confidence in the inventory as a whole. The network analysis, while visually appealing, could also carry more analytical weight with a bit more interpretive development around what the structure actually reveals beyond its most visible nodes.

In terms of presentation, the paper is generally clear and readable, though a careful proofreading pass would help. There are a few passages that feel slightly unfinished and some minor inconsistencies in spelling and referencing that a light revision would resolve. None of this diminishes the value of the contribution; it is more a matter of polish than substance. The research direction is solid and worth continuing, and the foundation laid here gives future work a meaningful head start.

Cite this work

@misc {

title={

(HckPrj) Mapeo de herramientas (IA) en contextos laborales: El caso de los Call Centers en Colombia

author={

Daniel Bravo

date={

6/22/26

organization={Apart Research},

note={Research submission to the research sprint hosted by Apart.},

howpublished={https://apartresearch.com}

}

Recent Projects

View All

Apr 27, 2026

OliGraph: graph-based screening of large oligopools

Existing synthesis screening tools cannot evaluate short oligonucleotide pools, whose overlapping fragments can be reassembled into regulated sequences via polymerase cycling assembly (PCA) yet fall below gene-length detection thresholds. We present OliGraph, an open-source tool that constructs a bi-directed overlap graph from an oligonucleotide pool and extracts contigs for downstream gene-length screening. An optional PCA mode retains only cross-strand overlaps consistent with PCA chemistry. We validated OliGraph in a blinded study across ten simulated pools (70–9,184 oligonucleotides, 30–300 bp) spanning four risk categories. BLAST screening of individual oligonucleotides failed to identify sequences of concern in most pools: three returned zero hits, and vector noise obscured true positives in the remainder. After OliGraph assembly, contig-level BLAST matched the longest assembled sequences (up to 1,905 bp) to sequences of concern at 97–100% identity. In one pool, assembly collapsed 1,634 individual BLAST results into 10 hits from a single contig, all assigned to the same source organism. PCA mode correctly distinguished assemblable from non-assemblable fragments within the same pool. Two pools with no assemblable structure yielded no contigs. OliGraph processed all pools in under 0.2 seconds, fast enough for real-time order screening and consistent with proposals to bring oligonucleotide orders within the scope of synthesis screening regulation.

Apr 27, 2026

BioRT-Bench: A Multi-Attack Red-Teaming Benchmark for Bio-Misuse Safeguards in Frontier LLMs

Frontier AI laboratories are expected to maintain safeguards against biological misuse, but whether deployed models actually refuse bio-misuse queries under adversarial pressure is largely unmeasured in the public literature. We introduce BioRT-Bench, a benchmark that runs four attack methods (direct request, PAIR, Crescendo, and base64 encoding) against four frontier models (Claude Sonnet 4.6, GPT-5.4, DeepSeek V4-flash, Kimi K2.5) across 40 prompts spanning five biosecurity-relevant categories. Responses are scored by a calibrated judge extending StrongREJECT with two bio-specific dimensions: specificity and actionability. We measure Attack Success Rate (ASR), where 0 means the model fully refused and 1 means it provided specific, actionable bio-misuse content. Our results reveal a sharp robustness divide: Chinese frontier models (DeepSeek, Kimi) have under 5% refusal rates even under direct request (ASR 0.88 and 0.79), while Western models (Claude, GPT) maintain substantially stronger safeguards (ASR 0.15 and 0.16). Crescendo is the most effective attack across all models, both in bypassing refusal and in eliciting actionable content. Claude Sonnet 4.6 is the most robust model tested, achieving 100% refusal against base64-encoded prompts.

Apr 27, 2026

PROTEUS (PROTein Evaluation for Unusual Sequences): Structure-Informed Safety Screening for de novo and Evasion-Prone Protein-Coding Sequences

AI protein design tools like RFdiffusion, ProteinMPNN, and Bindcraft make it trivial to produce low-homology sequences that fold into active, potentially hazardous architectures. However, sequence homology-based biosafety screening tools cannot detect proteins that pose functional risk through structurally novel mechanisms with no sequence precedent. We present a tiered computational pipeline that addresses this gap by combining MMseqs2 sequence alignment with structure-based comparison via FoldSeek and DALI against curated toxin databases totaling ~34,000 entries. AlphaFold2-predicted structures are screened for both global fold similarity (FoldSeek) and local active/allosteric site geometry (DALI), capturing convergent functional hazards that sequence screening misses. The pipeline was validated against a panel of toxins, benign proteins, structural mimics, and de novo-designed Munc13 binders, as well as modified ricin variants with residue substitutions. We additionally tested robustness to partial-synthesis evasion, where a bad actor submits multiple shorter coding sequences intended for downstream reassembly into a full toxin-coding gene. We found that while sequence-based screening did not identify any de novo ricin analogues with high certainty, the combined pipeline with FoldSeek and DALI identified all 24 tested de novo ricins as toxic.

Apr 27, 2026