Related articles
Insights to understand the work environment, anticipate risks, and make decisions with clarity, free from noise and unnecessary jargon.
View all
The conversation about artificial intelligence in business usually revolves around two poles: the excitement for the efficiency it promises and the concern for the jobs it might eliminate. Both perspectives, while understandable, are insufficient for those responsible for leading an organization or managing human capital. The real challenge is not whether to adopt AI—that decision, in practice, has already been made by the market—but rather to identify the specific operational risks its implementation creates and prepare to manage them before they turn into legal contingencies or losses in competitiveness.
This analysis addresses five of those risks that, by their nature, require direct attention from executives and human resources departments: the accelerated disruption of specialized professions, the problem of algorithmic opacity, ethical dilemmas in automated selection processes, the evidentiary value of AI-generated outputs in labor disputes, and the strategic urgency of reskilling existing talent.
For years, automation was perceived as a phenomenon that primarily affected operational workers: cashiers replaced by self-service terminals, robotic assembly lines, and logistics processes without human intervention. That stage has largely already occurred. The current frontier of technological disruption has shifted toward specialized professions, and that radically changes the risk profile for companies.
Software programming is perhaps the most visible example. Until very recently, mastering programming languages was practically a guarantee of employability and high income. Today, generative AI systems are capable of writing functional code, debugging errors, and proposing software architectures at a speed that no junior programmer can match. This does not mean that developers will disappear, but it does mean that the profile demanded by the market is mutating: it is no longer enough to know how to program; one must know how to direct, audit, and complement what AI produces.
Something similar is happening in the legal field. Current AI platforms can analyze entire bodies of law, cross-reference jurisprudence, identify relevant precedents, and generate drafts of legal documentation in a fraction of the time it would take a junior associate. This does not make the experienced lawyer dispensable—legal judgment, procedural strategy, and negotiation skills remain irreplaceable—but it does drastically reduce the need for man-hours dedicated to research and routine drafting tasks.
Medicine is moving in the same direction. Work is already underway on models where AI performs an initial symptom screening before the patient sees a doctor, which will inevitably reconfigure the roles of health professionals in corporate and occupational medicine settings.
For executives, the implication is clear: if your organization employs lawyers, accountants, financial analysts, programmers, corporate doctors, or any professional whose work involves data processing and data-driven decision-making, AI is already transforming what those roles require. Ignoring this transformation does not protect jobs; it simply delays the moment when the skills gap becomes unmanageable.
One of the most relevant concepts for understanding the risks of AI in the workplace is the algorithmic black box. In simple terms, it refers to the fact that in many artificial intelligence systems—particularly those based on deep learning—not even the people who designed the algorithm can explain with precision why the system reached a certain conclusion or recommendation.
This has direct legal implications. If a company uses an AI system to evaluate employee performance or to determine who receives a promotion, and an employee challenges that decision, the company will need to explain the criteria that led to the result. If the system operates as a black box, that explanation simply does not exist, which places the organization in a position of significant procedural vulnerability.
International principles on AI—both those from UNESCO and the European Union Act—insist on three requirements that point directly to this problem: transparency, traceability, and explainability. The algorithm must be auditable, its logic must be reconstructible, and its results must be justifiable to the affected employee and, if necessary, to labor authorities.
In the Mexican context, the regulation of digital platforms already incorporates this requirement through the Algorithmic Policy Management document, which obliges the employer to inform the worker about the operation of the algorithm that affects their working conditions. Although this obligation is currently limited to the realm of digital platforms, the regulatory trend suggests that it will progressively extend to any employment relationship where an automated system participates in decision-making.
The operational recommendation is straightforward: before implementing any AI tool in processes that affect workers, ensure that the provider can explain how the algorithm works, what variables it weighs, how it learns, and what mechanisms exist to audit its results. If the provider cannot answer those questions, the risk you are assuming is considerably greater than it appears.
One of the fastest-growing uses of AI in recruitment processes is the automated interview. In its most advanced version, the candidate connects to a video call where they interact not with a human, but with an AI-generated avatar that asks job-specific questions, analyzes responses in real time, and generates a candidate evaluation.
From an operational efficiency perspective, the tool is powerful: it allows for the screening of large volumes of candidates in significantly less time and with standardized criteria. From an ethical and legal perspective, however, it raises questions that companies must address before implementation.
The first issue is equitable access. Not all candidates have the same familiarity with digital environments or the same ability to perform in front of an artificial interlocutor. Some people, due to generational, cultural, or simply temperamental reasons, perform significantly worse in an interview with an avatar than in a conversation with another human being. If the system penalizes this performance difference without considering that it does not reflect the candidate's actual competence for the role, the result is a form of exclusion that, even if unintentional, can have legal consequences.
The second issue is the information the system captures beyond verbal responses. Some automated interview platforms analyze facial expressions, tone of voice, eye movements, and other biometric indicators to infer personality traits or levels of reliability. This type of sensitive personal data processing falls squarely into the realm of data protection legislation and, depending on how it is implemented, may constitute a violation of the candidate's rights.
The recommendation for human resources departments is not to adopt these tools without a prior analysis that considers both their legality under the Mexican regulatory framework and their real impact on the diversity and inclusion of the selection process. Technology can be extraordinarily useful, but its implementation must be accompanied by safeguards to ensure that equity is not being sacrificed for the sake of speed.
This question may seem premature, but the reality is that it is already being raised in Mexican forensic practice. Current AI systems are capable of generating technical reports, data analysis, and specialized assessments with a level of structure and reasoning that, at least on the surface, is comparable to that of a human expert.
However, the Mexican labor procedural system rests on principles that still require human intervention as a requirement for validity. For an expert report to have full evidentiary value, it must be issued by a licensed professional who can be called to ratify it before the court and who assumes personal responsibility for its content. An AI system meets none of these conditions.
This does not mean that AI cannot participate in the preparation of reports. In fact, it already does: it is a tool that allows human experts to process larger volumes of information, identify patterns that would otherwise go unnoticed, and structure their conclusions with greater analytical rigor. The difference is that the final product must bear the signature and responsibility of a natural person authorized to practice.
Now, there is an additional element worth considering. If a report is prepared with the support of AI, it is technically possible to document the entire conversation held with the system: the instructions given, the information provided, the responses generated, and the corrections introduced by the human expert. This traceability, far from detracting from the report's credibility, can strengthen it by evidencing the methodological rigor with which it was constructed.
The procedural horizon suggests that labor courts will gradually develop specific criteria for evaluating the role of AI in the production of evidence. Companies and legal professionals who transparently document the use of these tools now will be better positioned when those criteria are formalized.
Of all the topics linked to artificial intelligence in the workplace, one transcends the legal and regulatory to enter the realm of organizational survival: training existing talent.
AI will not replace a competent professional. What it will do—and is already doing—is make a person who masters AI tools substantially more productive, faster, and more precise than someone with the same technical skills but without that ability. In a competitive market, that difference translates into an advantage that companies cannot afford to ignore.
The concept of upskilling refers to expanding an employee's skills within their current role: a lawyer learning to use AI for legal research, a recruiter mastering algorithmic filtering tools, or a financial analyst integrating predictive models into their daily work. reskilling, on the other hand, involves retraining an employee so they can perform different functions than those they originally held, anticipating that their current position may be transformed or absorbed by automation.
For executives, investing in these programs is neither a discretionary expense nor a mere employee perk: it is a business decision with a direct impact on productivity, talent retention, and the organization's ability to adapt to a regulatory and technological environment moving at an unprecedented speed.
Companies that postpone this investment, arguing that AI is still in its early stages, are making a miscalculation. The technology that currently requires users to write precise instructions to obtain a useful result—known as prompting— will evolve toward increasingly intuitive interfaces, just as computing moved from text-based code to graphical interfaces and then to touch navigation. Each technological generation lowers the barrier to entry, but those who arrive first accumulate an advantage that widens over time.
The recommendation is to start now. You don't need expensive programs or sophisticated platforms to take the first step. Publicly accessible generative AI tools allow any professional to start experimenting, understanding the capabilities and limitations of these systems, and identifying concrete applications in their daily work. What is needed is an institutional decision to prioritize this training and integrate it into the organizational culture as a strategic competency, not just a technological curiosity.
Artificial intelligence is not exclusively a matter for the IT department. It is a matter for general management, human resources, regulatory compliance, and corporate strategy. Companies that approach it with this holistic vision will be prepared not only to comply with upcoming regulations but to turn the responsible adoption of AI into a genuine source of competitive advantage.
.png){kind=logo}