The conversation about technological innovation usually starts with promises: Generative AI, automation, cyber resilience, advanced analytics. However, as we always like to remember at MCPRO whenever we can, innovation is done by people and those people who innovate need the necessary knowledge. So the capital necessary for innovation is talent. And it is not just any talent, but the critical one, the one who knows our business, the flexible one, the adaptable one. And this is not found in word-based ads. What’s more, sometimes it is found in our own company. But let’s look at the data.
The numbers confirm the magnitude of the problem. According to recent data from OLD Spainless than half of digital initiatives achieve their business objectives, and the main barrier is not budget or technology: is the absence of practical capabilities. The global deficit of cybersecurity professionals is estimated between 2.8 and 4 million people according to various sources, with ISC2 placing it at 4.8 million and Fortinet reporting direct consequences: in Latin America, 86% of organizations suffered at least one cyber intrusion in 2024, while 329,000 security positions remain vacant in the region.
But the problem is not just that there is a lack of people. The thing is that we continue to think that the solution is only to hire more, when in reality the main lever is elsewhere: govern the talent you already have and orchestrate the talent you need from outside.
Talent governance: business knowledge is not hired, it is built
Before talking about governance and skills arbitrage, it is worth pausing on something that many organizations forget: Technical competence without business context is not strategic talent. A brilliant machine learning engineer who doesn’t understand why the billing system can’t go down on a Friday afternoon, or who doesn’t know the particulars of your industry’s purchasing process, is a partial resource. You can solve technical problems, but you can’t anticipate risks, prioritize correctly, or design solutions that fit the organization’s actual culture and constraints.
From this perspective, skill arbitrage is not a binary decision between internal and external, but a strategy for assigning work to the best available resource without losing control over critical knowledge. It’s about answering three questions before starting to hire. First, if this knowledge is strategic for our business. Second, if we touch sensitive data, operational continuity or compliance. And third, whether we have the time and capacity to develop this capability internally. When the answers to the first two are affirmative, ownership must be internal even if you collaborate with external specialists. When the third is negative, accelerated training and reconversion of internal profiles are more profitable than looking for unicorns in a saturated market.
The mix of capabilities: internal humans, freelancers, partners and AI agents
The most agile organizations no longer manage homogeneous workforces. They operate with a deliberate mix of four layers of capacity, each with its specific function.
Internal human capital is the guardian of business knowledge. Translate strategy into architecture, understand undocumented operational constraints, know the history of why certain processes are the way they are, and can anticipate which technological changes will generate resistance or acceleration. These responsibilities are your strategic core: architecture, security, sensitive data, product ownership. Their value is not only in what they know how to do technically, but in what they know about the business.
But here appears a new layer that deserves attention: the emerging roles of AI agent monitoring and optimization. Deloitte estimates that by 2026 organizations will begin significant investments in training programs for these specialized roles, recognizing that they represent a critical competitive differentiator.
We are talking about Agent-Ops teamsmulti-agent orchestration engineers and specialists in streamlining AI resource consumption. It’s not science fiction: IBM already reports saving more than 3.9 million employee hours through agent-based automation, but those savings require expert human oversight to maintain quality, optimize API costs, and prevent agents from consuming resources inefficiently.
Los freelancers y gig workers They provide speed and punctual expertise in peaks of demand. According to data from Malt, Spain has approximately 700,000 independent or freelance professionals in the digital ecosystem, a figure that has grown by 40% in the last decade. 64% of these professionals chose this modality due to a personal strategic decision, not out of necessity, which reinforces that freelancing is a mature professional option. They are useful for limited projects, emerging technologies that do not yet justify permanent internal roles or to cover temporary gaps while internal capacity is developed.
Partners and suppliers they industrialize and scale. When you need to deploy critical infrastructure, manage multicloud environments or implement security solutions that comply with complex regulatory frameworks, what you need is a partner with muscle, certifications and contractual commitment. The key is not to outsource everything, but to maintain internal governance over architecture and security while delegating operational execution.
Finally, AI agents They must now be considered as part of the productive mechanism. Automate repetitive tasks and They free up human capacity for work of greater strategic value. Intelligent agents can take on repetitive tasks, speed up investigations, and bolster analyst capabilities. They do not replace engineers, they multiply them. Code generation, documentation, log analysis, automated testing, first-level support responses, anomaly detection, quality monitoring or customer feedback integration are use cases where agents already work.
But as IBM warns, autonomy does not mean eliminating human supervision: AI teams need to invest in evaluation, reliability, efficiency optimization, scalability and maintainability so that these systems generate real value and do not become a source of problems.
From the ticket factory to the product organization
The most significant change It’s not technological, it’s conceptual.. IT is changing from being a factory of incidents and fire equipment to becoming a product organization. In this model, upskilling and reskilling are no longer optional and become a strategic investment. As we already mentioned in a previous article about upskilling and reskilling, companies that invest in internal academies and retraining plans not only retain better talent: they develop it. Because knowledge is distributed and stops depending on three key people who, if they leave, leave irrecoverable gaps.
The difference is that this upskilling does not seek to make everyone a specialist in everything, but rather build layers of distributed competence: advanced digital literacy for all IT (data, security, automation, cloud, AI fundamentals), specialists for critical domains (MLOps, cloud security, data engineering) and translators between business and technology (product managers, analytics translators, security champions). The objective is that critical knowledge is not concentrated in unattainable profiles.
The first is to declare the vision and measure it. “No one is left behind” It is not a slogan, it is a competitiveness policy with indicators: percentage of converted roles, effective internal mobility, relevant certifications obtained, average upskilling time applied to real projects. And with the very high rate of innovation that we are experiencing, retaining personnel who continue to know the business and the functioning of the department, are immersed in the company culture and have the flexibility to train and keep up to date is essential, both for the future of the company and for your professional future.
The second move is to design a future role architecture. You need to know what roles grow in the next three years. MLOps, cloud security, data engineering and AI applied to processes are the obvious ones, but completely new categories appear that many organizations still do not consider. Agent-Ops teams, multi-agent orchestration engineers, and AI resource consumption optimization specialists They emerge as critical roles.
Gartner projects that by 2026 approximately 40% of enterprise applications will incorporate task-specific AI agents, up from less than 5% in 2025. This map allows you to anticipate investments in training before the need is urgent.
The third step is create an academy connected to real products. We are talking about modular training with projects in which learning is validated with deliverables, internal mentors who transfer practical knowledge and evaluation by results, not by attendance. The academies that work are not classrooms but laboratories where you learn by doing.
The fourth move is to implement a product operating model. Domain teams, agile practices integrated with DevSecOps, clear ownership of measurable results. The product model is not just for startups: it is the way IT organizations go from reactive to proactive, from executors to co-creators of value.
The fifth and last is to govern hybrid human plus AI work, create policies and manuals of good practices aligned with the business culture. You need to define what data they can process, how privacy is protected, who owns the intellectual property generated. Mandatory human review on critical decisions, traceability of what the agent did and what the person did, and security controls that prevent a compromised agent from becoming a backdoor.
But you also need operational governance: who monitors the resource consumption of agents, how their efficiency is optimized, what metrics determine whether an agent is providing value or only generating cost. As CGI warns in its Human-Agent Collaboration Framework, AI agents need clear rules, strong metrics, and ethical oversight that ensure results that are reliable, auditable, and aligned with business values.
The new contract: learning is part of the job, not an extra
The most important cultural turn is accept that training is no longer a benefit or an add-on: it is a core competency of the business. The organization that wins is not the one that hires the fastest, but the one that retools skills better, more often, and with less friction. And this requires a change in the psychological contract between company and professional: continuous development stops being an individual responsibility and becomes an organizational system.
In that context, The “no one left behind” promise is not just talent retention: is active capacity building. And skill arbitration is not downsizing or precariousness: it is operational intelligence to place each person, each freelancer, each partner and each AI agent where they contribute the most value without losing governance over critical knowledge.
Because in the end, innovation is not implemented by purchasing technology. It is enabled by building a system where people evolve with the business, structures adapt to real needs and technology amplifies capacity rather than replacing it. And that, in a world where the shortage of talent in AI, cybersecurity and data will continue to be a limit for many companies, is the difference between truly transforming or getting stuck in the presentation of the digital strategy.
