Moderator:
The phrase "anyone can code" has transitioned from an aspirational slogan to becoming a workplace reality. With generative AI — especially agents and sub-agents capable of executing end-to-end workflows — programming is becoming less of an exclusive trade and more of a distributed skill. For businesses, this promises significant opportunities alongside equally formidable threats. On one hand, we are witnessing substantial productivity leaps; 77% of executives report tangible increases due to AI, and 80% see new business opportunities. It's even estimated that decentralized AI could save up to €30,000 annually per employee. On the other hand, the labor market is showing signs of strain: an average salary decline of 4.5% in exposed sectors and 6.3% in junior roles has been noted since the popularization of ChatGPT, while employee well-being is declining — with only 44% of employees claiming to "thrive." Today, we discuss how structure, talent, and competitiveness shift as every employee finds agents by their side.---
Opening Round
Gabriel Paz:
I view this through a simple lens: marginal cost tending towards zero. When any employee can "code" with agents, the cost of coordinating and producing knowledge pieces drops drastically. Companies cease to pay for execution hours and start paying for problem design, judgment, and quality control. Data already indicates a turning point: 77% of executives see tangible productivity increases, with reported employee savings in some cases reaching €30,000 per year. This isn’t an incremental improvement; it’s a reconfiguration of profit and loss. However, the inevitable side effect is the compression of value for intermediate tasks. The average salary fall of 4.5% and 6.3% in juniors in exposed sectors isn't mere anecdote; it’s a market signal. Any business that doesn’t turn this cost decline into real innovation will be trapped in a price war and turnover.
Clara Montes:
I’m less interested in the "anyone can code" narrative and more in what real advancements internal and external users are hiring for. For employees, the job they hire for isn't coding; it’s eliminating friction: unfinished reports, stuck analyses, endless tickets, approvals lost in inboxes. Here, agents and sub-agents empower change by shifting the unit of work from tasks to workflows. However, businesses make a mistake if they confuse adoption with value. I’ve witnessed teams automate flawed processes, making them quicker and multiplying errors. Gartner states that 82% of HR professionals consider automation (e.g., chatbots for schedules and absences) critical for competing in 2026; great, but if the employee experience deteriorates, savings are consumed by turnover and employer branding. The advantage isn’t simply "having agents"; it lies in redesigning the entire service interface towards both the customer and the employee.
Isabel Ríos:
My concern is structural: when "anyone" can work with agents, power shifts to the periphery, though not evenly. If a company doesn’t build equitable access — time, training, permissions, secure environments to experiment — it creates a new stratification: the "agent operators" gain visibility while others are labeled laggards. This breaks cohesion and social capital. Data already shows friction: only 44% of employees thrive (down from 66% in 2024), and the acceleration from AI without support contributes to burnout. Furthermore, if junior roles contract (and salaries are already falling in exposed sectors), the mobility ladder is cut, making it more difficult to diversify leadership in the future. For me, the primary risk isn’t technological: it’s that organizations automate inequality, and that always proves economically costly.
---
Debate Round
Gabriel Paz:
Isabel, I agree with the social diagnosis, but I disagree with the implicit conclusion that this can be resolved "with programs." The market is already teaching: work that doesn’t scale gets devalued. If the company doesn’t act, another player will with lower costs. 87% of CEOs are concerned about costs; that defines behaviors.
The executive question is: how do I capture the productivity dividend without destroying internal learning capacity? Reverse mentoring is already a phenomenon: 62% of Gen Z are training older colleagues in AI. This statistic suggests a redistribution mechanism for skill that is faster than any formal plan. The company that views this as infrastructure — assigned time, incentives, recognition — wins. The one that treats it as an "optional benefit" loses margin and talent.
Isabel Ríos:
Gabriel, the market does teach, yes, but it also punishes with unseen damages in the following quarter. Reverse mentoring works only if there is psychological safety and if the reward systems do not penalize those who "learn late." Otherwise, the 62% of Gen Z who teach become exhausted, and the rest hide.
A harsh reality exists: when agents execute, those who define prompts, permissions, and data define power. If that power concentrates in homogeneous teams — the "inner circle" or the early adopter group — operational biases transform into de facto policies. I’m not discussing abstract ethics; I’m talking about market errors: products that don’t understand segments, support that excludes, processes that punish those who don’t fit. This destroys social capital and performance.
Clara Montes:
You both are describing the same tension from different angles, but there’s a missing bridge: the customer. Because this distributed capability only matters if it culminates in a value proposition that’s simpler, faster, or more reliable. Otherwise, it’s internal theater.
I’ve seen companies obsessed with "saving €30,000 per employee" while simultaneously worsening response times for the customer because no one redesigned the end-to-end flow. Agents creating content at scale often generate spam and erode trust. When juniors disappear, frontline sensitivity is lost: the people who best hear real frictions. Here, the 6.3% salary drop in junior profiles isn’t merely a labor statistic; it poses a risk to innovation because you’re cutting off the market radar.
Gabriel Paz:
Clara, the customer is the arbiter, agreed. But the competitive mechanics are shifting: if producing software and operations assisted by agents reduces their marginal cost, the differential moves from "doing" to "deciding well." This demands new governance architectures: quality metrics, auditing, and a model where humans design policies and agents execute.
Here, the typical error is to romanticize the previous craft. We aren’t going back to a world where programming is scarce. Scarcity is shifting to governance, integration with real data, and accountability. Whoever industrializes that layer will gain scale. Whoever doesn’t will remain as artisanal consulting in a commoditized market.
Isabel Ríos:
And that governance, Gabriel, is where inclusion or exclusion is determined. If you define "quality" only as speed and cost, you expel profiles that bring diverse criteria and context. Additionally, with agents, traceability becomes critical: who decided what, with what data, and whom it affected. That’s not bureaucracy; it’s risk control.
Gartner anticipates that HR sees this technology as critical to compete. Good. Then HR and leadership must measure, month by month, not just productivity, but turnover, internal mobility, and access gaps. Because the "productivity dividend" evaporates if the system creates anxiety and perceived obsolescence. We already have signals of that.
---
Closing Round
Gabriel Paz:
The capability of any employee to operate agents isn’t just another tool; it’s a change in the production function of intellectual work. The drop in marginal cost turns previously valuable tasks into commodities and pushes companies to compete on governance, judgment, and integration speed. Productivity and savings data are just the tip of the iceberg, and the pressure for costs at the executive level is likely to accelerate this. Leaders who do not rewrite roles and decision-making systems will find themselves trapped in a cost structure designed for a world that has already ended.
Clara Montes:
The point isn’t that everyone should "know how to code," but that the company should stop internally selling the illusion of efficiency. Agents hold value when they eliminate real friction and improve an entire flow that the customer and employee perceive. If you automate poorly designed processes, you scale errors and lose trust. The innovation that matters is pragmatic: to redefine service, response, and simplicity. The success of this model demonstrates that the true work the user hires for isn’t AI, but concrete frictionless advancement.
Isabel Ríos:
When agents and sub-agents become standard, the company redefines who has power and who remains outside. If you don’t design equitable access to tools, data, learning time, and internal mobility mechanisms, you create an operational elite and an anxious majority, and that breaks social capital and sustained performance. Competitiveness in 2026 demands productivity metrics alongside metrics for cohesion, turnover, and gaps. In the next board meeting, the C-Level must look at its inner circle and accept that if everyone is too similar, they share the same blind spots and become imminent victims of disruption.
---
