10x AI Professionals represent the dawn of a new work paradigm, where Artificial Intelligence not only multiplies the productivity of engineers but also empowers every professional. Far from being a myth limited to the software industry, the concept of 10x AI Professionals opens the door to a profound transformation of the future of work, where technology acts as a strategic copilot to amplify human potential.
The software industry is undoubtedly continually seeking methods to accurately measure developer productivity. Initially, the focus was on the number of lines of code (LOC) produced by each engineer. However, it soon became clear that quantity did not necessarily equate to quality or efficiency.
Consequently, the industry evolved toward more sophisticated metrics, such as story points, submitted tickets, and accepted user tests. These metrics aim to provide a comprehensive view of productivity by considering the complexity and outcome of completed tasks.
The ultimate goal has always been to identify the most effective team members. By identifying these key contributors, organizations hope to improve team performance and encourage best practices across the workforce. The constant quest to accurately measure productivity reflects the dynamic nature of the software development industry and continues to evolve.
Consider the «10x engineer» or «10x developer,» a term suggesting that some engineers are ten times more productive than their colleagues. This concept has generated much debate and controversy in the industry. These individuals are extremely bright and can accomplish in one hour what others would take ten hours to complete. They are highly skilled at fixing major bugs and bringing the system back online if it goes down.
Recently, a Harvard/BCG study estimated that consultants could complete 12% more tasks and finish them 25% faster with GPT-4. This was just the average using that technology. There are important discussions regarding the profound impact of AI on the labor market and jobs for at least the next five years
By 2030, AI is expected to automate millions of jobs. However, it will also create new opportunities, particularly in AI development, maintenance, ethics, and training. Repetitive and basic analysis jobs are expected to disappear, while new professions will emerge, such as prompt engineer, AI-assisted healthcare technician, and AI ethics officer.
However, there is no universal scale for measuring productivity, and the concept must be relativized. In a managed environment, the gap between most software engineers is relatively modest, and the performance of the same person across multiple tasks varies. Individuals can outperform their peers in some areas and perform averagely in others. Furthermore, given the current rise of AI, increased productivity would not be automatic or immediate. It would depend on fine-tuning the characteristics of the tool, task, and professional context.
From Engineers to 10x Professionals
Beyond questioning the scale of the 10x engineer, this concept is an interesting trigger for thinking about productivity in other professions. While it doesn’t yet appear that there are 10x professionals in areas such as marketing, talent recruitment, or financial analysis, some experts assume that as more jobs become integrated with AI, this category of 10x engineers could be extrapolated to 10x professionals.
Why hasn’t this happened yet? There are still no 10x professionals because, in many positions, the gap between the best and the average worker is limited. No matter how athletic a supermarket cashier is, they are unlikely to scan products fast enough to get customers out of the store 10 times faster. Similarly, even the best doctor is unlikely to get their patients to recover 10 times faster than the average doctor—but for a sick patient, even a small difference is very important. In many jobs, the laws of physics limit what any human or AI can do, unless we completely reinvent that job.
But for other jobs that primarily involve applying knowledge or processing information, AI could impact the task. This happens because in some positions, people with technological knowledge emerge who are supposed to coordinate a set of technological tools to do things differently and begin to have, if not 10 times the impact, easily twice as much.
Although 10x engineers don’t write code 10 times faster, they make technical architecture decisions that result in significantly better downstream impact, detect problems and prioritize tasks more effectively, and instead of rewriting 10,000 lines of code (or labeling 10,000 training examples), they could figure out how to write just 100 lines (or collect 100 examples) to get the job done.
Extending this to other professions—possibly traditional marketers, recruiters, and financial analysts—would do things differently in this new scenario. For example, perhaps traditional marketers repeatedly write social media posts. 10x marketers could use AI to help with writing, but the transformation will be more profound. If they are deeply sophisticated in how they apply AI (ideally able to write code themselves to test ideas, automate tasks, or analyze data), they could end up conducting many more experiments, gaining deeper insights into what customers want, and generating much more precise or personalized messages than a traditional marketer, thus having 2, 5, or 10 times greater impact, depending on the case.
Similarly, 10x recruiters won’t just use generative AI to draft candidate emails or summarize interviews. They might coordinate a suite of AI tools to efficiently identify and research a broad pool of candidates, allowing them to have a far greater impact than the average recruiter. And 10x analysts won’t just use generative AI to edit their reports. They might write code to orchestrate a set of AI agents that conduct in-depth research on products, markets, and companies, yielding far more valuable insights than someone doing traditional research.
The Argentine Case: AI Pods as a Laboratory of the Future
Particularly in Argentina, a growing branch of engineering is the software and IT services industry. In this case, a global software company inaugurated the concept of «AI Pods,» which promises to revolutionize the software market: it allows companies to do the work of five experienced professionals (5X engineers), with the flexibility offered by a Netflix-like subscription model.
Each pod operates with a token system that defines its capacity, powered by AI workflows overseen by human experts, promising measurable and efficient results for each client and reducing the time required to update a system’s code by one-fifth.
Although AI Pods work with autonomous agents, human oversight is still essential, especially in sensitive industries such as banking and healthcare, where current regulations and standards must be strictly adhered to.
Techno-optimists argue that as AI becomes more useful in many more jobs, similar paths will open up for many more people to become «10x professionals».
However, we must ask ourselves: what factors would make it possible to talk about 10X professionals at the local and regional level? What ingredients of human creativity do these tasks have, beyond hard and technical skills? And, above all, does doing something faster and with fewer people necessarily mean the result will be of higher quality?
Problems that the notion of 10x professionals may bring with the arrival of AI
As AI increasingly penetrates the labor market, problems will arise related to task dynamics and the knowledge required to perform them efficiently. For example, there will be questions about the degree of autonomy that AI agents can have with respect to humans, as well as the training that these systems, especially LLMs, will require to better understand the work environment.
In this sense, we list at least five problematic issues to be resolved:
1) Artificial intelligence agents will function best in environments prepared for this type of development, which are standardized platforms.
If processes have thousands of rules, if an AI agent has to help a human agent pay pensions and retirement benefits from the pension system, or life or unemployment insurance, the problems will surely begin there. Because these are not highly standardized systems and infrastructures.
The trend toward homogenization and standardization of tasks is just around the corner, but in areas such as the public pension system or large systems like health insurance, nothing is yet resolved or standardized.
2) These standardized platforms always present a latent tension in the degrees of freedom of the product being built, depending on whether the demand involved in the task is more rigid or flexible.
Imagine a marketing professional who writes copy or advertisements. Surely, if they delegate this task to an AI agent to do it much faster, prepare a guide of questions with ordered answers, and the system will be able to do it if the copy typically written is more or less standardized and structured, it will adapt well.
Now, if you have a heterogeneity of input material, a variety of advertisements and commercials depending on the client, and the agent needs to be a little more flexible regarding the content, it will be a challenge to categorize it and do it more quickly.
Let’s also consider the work of a human resources recruiter who must select for different positions, analyze, and evaluate different professional profiles. Will it be a highly standardized task or rather heterogeneous, given the variables to be considered: skills and competencies, salary expectations and job-specific characteristics, and the attributes that each optimal selection takes into account for each position?
Everything depends on the level of creativity and communication skills required in the search. Simply assuming that jobs such as software development, personnel selection, marketing, or copywriting don’t require high levels of human creativity or communication skills would be a somewhat forced, uncritical, and reductionist approach.
3) Developing a technical task faster doesn’t necessarily mean being more efficient and productive.
When we talk about 10X engineers, we’re talking about speed and productivity—that is, how this multiplication of productivity impacts the final software development—but it’s not necessarily a measure of quality and efficiency. Other metrics or standards of quality and efficiency need to be sought, for example, story points, tickets submitted, accepted user tests, and incremental innovation or improvements compared to previous versions.
Success is an abstract concept in itself, which can depend on a large number of factors, primarily human, soft, and learning factors.
Therefore, these types of engineers—although this could also apply to marketing professionals, recruiters, and analysts—should possess programming and architectural skills, a focus on the big picture and results, excellent communication skills, and a human-centered empathy. They should be problem solvers (thinking «outside the box» and finding intuition outside of their direct experience to find the best way to solve a given problem) and they should be mentors of young developers and lifelong learners.
4) Better tools don’t necessarily mean faster. Recent indicators have shown that AI slows down experienced programmers.
An experimental study conducted by the independent METR (Model Evaluation and Transparency Research) laboratory in the first half of 2025 challenges certainties. Tested under real-world conditions on open source repositories they know well, experienced developers were on average 19% slower when using generative AI tools than when not.
This unexpected performance is especially striking because the participants themselves mistakenly believed that AI had helped them go faster.
To obtain reliable conclusions about the real impact of generative AI on software development, METR recruited 16 developers with several tens to hundreds of hours of experience with language models and who have contributed, on average, for the past five years to mature open source projects (22,000 stars on GitHub, over one million lines of code).
The methodology of their randomized controlled trial (RCT) is based on a simple principle: randomly allow or deny the use of AI (primarily Cursor Pro with Claude 3.5/3.7) on a set of 246 technical issues that are part of their professional routine (bugs, refactoring, feature enhancements), and then measure the resolution time.
Work times were self-reported and supplemented by screen recordings. Deliverables were qualitatively assessed through submitted Pull Requests.
Despite their experience, the developers anticipated a 24% speedup thanks to AI. At the end of the experiment, this estimate decreased slightly but remained optimistic: a 20% perceived time savings.
In reality, tasks performed with AI assistance took an average of 19% longer than those performed without it, a difference not only contrary to participants’ expectations but also contrary to predictions made by external experts in the fields of economics (-39%) and machine learning (-38%).
METR identifies five main factors that could explain this delay: A) An imperfect use of the tools, especially with overly simple prompts; B) A still limited familiarity with AI interfaces such as Cursor; C) High quality standards in the projects studied, sometimes incompatible with the suggestions generated; D) Insufficient coverage of complex cases by the models; and E) A form of cognitive distraction related to experimentation with AI.
Far from concluding that AI impairs the performance of all developers in all contexts, the study primarily highlights what we noted earlier: that productivity gains are neither immediate nor automatic: they depend on a fine-tuning of the tool, the task, and the professional context.
5) The integration of professional work with AI agents raises more questions than definitive answers.
Among these questions are how the productivity of work performed is measured using standards such as quality or efficiency, whether it is better or worse to compare a traditional model of a task with an AI-assisted model, and whether other development areas have similar conditions to those of software engineering.
Specifically, we ask ourselves: what conditions will be necessary and sufficient for the theory of 10x engineers to be extrapolated to other professions, especially marketing, human resources, and financial analysis? Will it be possible, with the rise of AI, to speak of 10X professionals?
In closing
AI has enormous potential to transform software development and engineering by amplifying developers’ capabilities. Although AI cannot replace 10x developers entirely, it can enhance their productivity and innovative potential. The future points toward a collaborative synergy in which AI tools and human developers work together to create a new paradigm of efficiency and creativity in software and IT service development.
Finding and hiring a 10x engineer — or any 10x professional, for that matter — shouldn’t be the primary objective. Fortunately, the optimal strategy for most companies is to avoid constantly searching for the «best» engineers and developers in the hope of discovering exceptional talent. The focus should be on ensuring that competent engineers perform to their full potential and can be trained and encouraged to pursue lifelong learning, curiosity, and innovation. This will allow them to achieve personal and professional development metrics.
Rather than aspiring to individual excellence, the challenge is to aspire to excellent overall team performance. Organizations that transform ordinary teams into high-performing ones will surely be the winners, along with their 10x engineers.
As more jobs become integrated with AI, we must keep pace with this complex transformation by fostering new learning and adaptability to changing environments, tools, and the critical skills required for each new position.
Will 10x professions be a new scenario enabled by artificial intelligence? We will likely have to wait a few years and observe how the labor market evolves with the proposed AI-driven transformations to answer this question definitively.
