How artificial intelligence can do in seconds what many unions have spent decades turning into assemblies, pamphlets, strikes, positions, trips, budgets, and inflammatory speeches.
Artificial intelligence can revolutionize labor negotiations by replacing part of traditional union intermediation with systems based on data, transparency, auditing, and direct voting. Instead of depending on union bureaucracies, workers could use AI platforms to compare real earnings, identify unfair account deactivations, evaluate regulatory proposals, receive legal guidance, and negotiate better working conditions. This article examines that future through the case of app drivers, showing how AI can transform worker representation into something faster, more objective, and more efficient. Less grandstanding. More evidence. Fewer professional intermediaries. More direct power for workers.
Maurício Pinheiro
The dynamics of negotiation between employers and employees have always been complex. On one side, workers seek fair pay, predictability, safety, dignity, and better working conditions. On the other, companies seek efficiency, productivity, cost control, flexibility, and survival in increasingly competitive markets. Between these two poles, unions emerged: organizations created, in theory, to collectively represent workers.
In theory.
Union history has real merits. It would be intellectually dishonest to deny that unions played important roles in struggles for working hours, safety, minimum wages, vacations, protection against abuse, and basic labor rights. The problem is that human institutions rarely remain pure after discovering three things: power, budget, and bureaucratic stability.
The union was born as a tool. In many places, it became a career. It was born as the voice of the worker. In many cases, it became the voice of the union leader. It was born as a defense against employer abuse. In many environments, it became a machine of ideological liturgy, a bureau of compulsory intermediation, and political theater in which real workers serve as extras.
The central question of this article is simple: if the legitimate function of a union is to represent interests, organize information, reduce asymmetries, negotiate collectively, and monitor abuse, why could an increasing share of these functions not be performed more efficiently, transparently, and personally by artificial intelligence systems?
The short answer is: it can.
The long answer is: it can, it should, and it probably will — not because AI is morally superior, but because it does not need a sound truck, does not collect union dues to finance pamphlets, does not turn assemblies into ideological rituals, and does not need to pretend that every labor demand is always the preliminary stage of a world revolution.
The Old Problem: The Intermediary Who Discovers He Is Indispensable
The traditional union was born to solve a real problem: the individual worker is weak before the company. An isolated employee negotiates in fear. An isolated driver complains and is ignored. An isolated factory operator reports abuse and can be replaced. Collective organization corrects part of this imbalance.
So far, there is nothing to dispute.
The problem begins when the channel of representation develops interests of its own. The worker wants wages, safety, and predictability. The union leader wants relevance, a mobilized base, narrative control, permanent negotiation, contributions, political presence, and, in some cases, the comfortable feeling of being indispensable.
The consequence is predictable: mediation begins to live not from solving conflicts, but from continuously administering them.
A negotiation resolved simply, transparently, and quickly is excellent for the worker. But it may be terrible for someone who built an entire career explaining why negotiation must be slow, ritualized, collective, politicized, and, naturally, conducted by him.
This is where artificial intelligence enters as a mortal threat to bureaucratic unionism: it can transform what used to be a monopoly of intermediation into an automated, auditable, comparable service available 24 hours a day.
Workers Do Not Need Sermons. They Need Data.
A large part of labor negotiation suffers from an elementary problem: lack of reliable information.
What is the real average salary for a given position?
What is the company’s financial margin?
What would be the impact of a 5% payroll increase?
What is the risk of layoffs if the company loses competitiveness?
Which benefit has the highest perceived value for workers?
Which schedule reduces fatigue without destroying productivity?
Which performance metric is fair, and which one is merely a managerial fantasy dressed up as a spreadsheet?
Traditionally, these questions are handled at negotiation tables contaminated by rhetoric. The company says it cannot pay. The union says it can. The company presents selective numbers. The union presents selective outrage. And the worker, who should be at the center of the process, often becomes the audience.
A well-designed AI system could do something more useful: gather data, compare scenarios, simulate proposals, estimate impacts, and present verifiable alternatives to all sides.
This is not about replacing rights with algorithms. It is about replacing theater with information.
A labor-oriented AI system could analyze anonymized payroll data, productivity by sector, inflation indexes, labor market data, turnover, accidents, absenteeism, internal complaints, court decisions, previous collective agreements, benefit costs, and workers’ actual preferences. It could then generate negotiation proposals under different scenarios: conservative, moderate, and aggressive.
Instead of a three-hour assembly to approve a slogan, workers could receive a clear simulation:
“If the priority is an immediate wage increase, the expected gain is X, but the risk of job reduction is Y.”
“If the priority is flexible scheduling, the financial impact is lower, but projected satisfaction rises by Z.”
“If the priority is food benefits, the cost to the company is lower than a linear wage increase, and the net gain for lower-income workers is higher.”
“If the priority is protection against algorithmic dismissal, the recommended clause requires auditing, explanation, and human review.”
This does not eliminate conflict. But it eliminates a great deal of performance.
The New Frontier: Algorithmic Management
This is no longer fiction. Algorithmic management is already at the center of contemporary work. Digital platforms assign rides, calculate prices, evaluate performance, rank delivery workers, deactivate accounts, and define incentives through automated systems. Traditional companies also use algorithms to schedule shifts, measure productivity, monitor performance, and evaluate workers.
The OECD already treats algorithmic management as a reality made up of tools that instruct, monitor, and evaluate workers. The European Union has approved rules for platform work requiring greater transparency and preventing major decisions, such as dismissal or deactivation, from being made exclusively by automated systems. The ILO has published global case studies showing that AI and algorithmic management have already entered the center of social dialogue.
In other words, while certain unions are still debating whether assemblies must be held in person because “the human warmth of struggle” is irreplaceable, employers are already using digital systems to organize work at scale.
The irony is delicious: the union that resisted electronic voting for years now discovers that it must negotiate against algorithms. The leader who feared losing control of the assembly to an app may now have to face a predictive model trained on millions of operational data points.
Technology did not ask permission. It entered.
App Drivers: The Laboratory of the Future
App drivers are perhaps the best example of this new tension. They want better income, predictability, protection against unfair deactivations, transparency in fares, safety, and less abuse by platforms. At the same time, many value autonomy, flexibility, and the possibility of working without a direct boss.
The traditional union solution is usually predictable: classify, regulate, represent, charge, compel, standardize.
But the app driver is not exactly the industrial worker of the twentieth century. He may work full-time, part-time, occasionally, informally, professionally, or temporarily. He may work for multiple platforms. He may reject rides. He may turn the app on and off. He may want protection against abuse without wanting to become a classic employee. He may want collective negotiation without accepting a union structure that speaks in his name without actually listening to him.
In Brazil, the attempt to regulate app drivers through PLP 12/2024 revealed precisely this difficulty: how can protection be created without destroying the flexibility that many workers value? How can social security contributions be guaranteed without turning the platform into disguised formal employment? How can one represent a category that does not always see itself as a category?
Here, AI could play a far more sophisticated role than the traditional union.
Imagine an independent, auditable platform oriented by artificial intelligence, funded by a minimal voluntary contribution — an expression that, by itself, should already cause shivers in certain union bureaucracies — or by public funds for labor innovation.
It would not exist to speak on behalf of app drivers, but to return to them what they so often lack: clear information, comparable data, and real decision-making power.
In this new model of digital labor negotiation, each driver could track in real time his income, risks, rights, and the concrete impact of each regulatory proposal. Less speech. Less pamphlet. Fewer empty assemblies. More data.
Such an AI platform could, for example:
Compare net earnings by region, time, and platform: The driver would not see only how much entered the app, but how much actually remained after fuel, maintenance, insurance, idle time, empty trips, and platform fees. A ride that looks good at first glance could turn out to be terrible in the pocket.
Identify unfair deactivations: Vague messages such as “violation of community guidelines” could be confronted with collective data. If many deactivations happen after drivers reject unprofitable rides, complain about the platform, or work in certain areas, the case stops being isolated and becomes evidence.
Record complaints with evidence: Screenshots, receipts, routes, times, amounts paid, cancellations, and platform messages could be automatically organized. Less venting lost in WhatsApp groups. More useful history for negotiation, complaints, or legal action.
Simulate the economic impact of regulatory proposals: Before supporting a new law, minimum fare, mandatory contribution, or change in employment status, drivers could evaluate the effects on income, demand, costs, and flexibility. Many proposals sound beautiful in speeches but expensive in practice.
Vote digitally on real priorities: Drivers could choose directly what matters most: fare transparency, protection against deactivations, insurance, social security, safety during night rides, lower platform fees, or access to platform data. The will of the category would appear without needing to pass through the filter of half a dozen enlightened intermediaries.
Negotiate minimum standards based on aggregated data: Instead of slogans about “better working conditions,” negotiation would deal with concrete metrics: minimum value per kilometer, minimum value per minute, deactivation rules, response time for appeals, dynamic pricing criteria, safety, and insurance.
Receive personalized legal guidance: AI could explain, in simple language, which documents to keep, when to appeal, what type of complaint to file, and what risks exist in each path. It would not replace lawyers in complex cases, but it would reduce dependence on intermediaries.
Understand dynamic pricing: Not every “high demand” area means real profit. Sometimes the driver moves to a supposedly hot zone, spends fuel, loses time, and discovers that the bonus looked better on the map than in the wallet. AI could help calculate whether the ride is actually worth it.
Evaluate whether a regulatory change improves income: The platform would show whether a proposal truly increases drivers’ net earnings or merely creates new intermediaries, new positions, new meetings, new contributions, and new specialists explaining to workers why they must continue being represented by someone.
Taken together, this would be a new form of labor negotiation with artificial intelligence: more data, more transparency, and more direct power for the driver. It would not be representation based on grandstanding, but on evidence.
And perhaps that is exactly why this idea is so dangerous to traditional unionism.
Does anyone really believe that a company like 99 in Brazil, controlled by China’s DiDi — Uber’s global rival and the legitimate child of platform capitalism with a Mandarin accent — is not already doing part of this internally? Ah, communist China… always reminding us, with delicate historical irony, that even real socialism discovered the virtues of data mining, algorithmic efficiency, and market optimization when the subject is urban transportation.
The question, therefore, is not whether the data already exist. They do. The question is who controls them, who interprets them, and who benefits from them.
Today, digital platforms use artificial intelligence, predictive analytics, and algorithmic management to optimize rides, prices, incentives, deactivations, routes, schedules, and user behavior. The driver, on the other hand, often continues negotiating in the dark, receiving only the final result of decisions he does not see, audit, or fully understand.
An independent labor AI platform would reverse that logic. It would not eliminate collective bargaining; it would make it smarter. It would not replace worker defense; it would replace inefficient, opaque intermediation frequently captured by political interests. Instead of representation based on grandstanding, we would have representation based on data.
And perhaps that is exactly why this idea is so dangerous to traditional unionism.
Because, deep down, what frightens certain structures is not artificial intelligence replacing the worker. What truly terrifies them is the possibility that artificial intelligence may show the worker that he may never have needed so many professional translators of his own reality.
Hollywood: When Even Artists Had to Negotiate with AI
Recent Hollywood cases show that AI has already become a central issue in labor negotiations. The Writers Guild of America included protections regarding generative AI in its 2023 agreement. SAG-AFTRA, representing actors and performers, fought intense disputes over digital replicas, synthetic voice, consent, and compensation. In 2025, the video game sector agreement included rules on the use of digital replicas, consent, and disclosure.
These cases are important because they show two things at once.
First: AI is already powerful enough to concern highly qualified creative workers. We are not talking only about cashiers, forklift operators, or drivers. We are talking about screenwriters, actors, voice actors, and professionals whose work involves identity, expression, and authorship.
Second: unions were only effective when they treated AI concretely: clauses, consent, compensation, limits, transparency, and auditing. Not when they made generic speeches against “technology,” but when they negotiated specific terms.
That is the lesson: the real struggle is not “against AI.” The real struggle is for an institutional architecture capable of using AI without turning workers into disposable raw material.
And the best architecture may not be the traditional union. It may be a combination of AI, auditing, direct voting, smart contracts, flexible representation, and data-driven negotiation.
The New York Times Case: When AI Becomes the Worker’s Inspector
In 2026, unions connected to The New York Times accused the company of using AI to monitor employees and violate collective agreements. The accusation is relevant because it reveals the darker side of technology: AI can be a channel for negotiation, but it can also become an instrument of surveillance.
This point must be faced honestly. AI is not an angel. Nor is it a demon. It is an amplification machine. It amplifies efficiency when designed to solve problems. It amplifies abuse when designed to control people. It amplifies justice when subjected to auditing. It amplifies arbitrariness when it operates as a black box.
Therefore, replacing the union with AI cannot mean handing the worker over to the employer’s algorithm. That would merely replace the assembly bureaucrat with a silicon bureaucrat — and, admittedly, with less coffee and more opacity.
The proposal here is different: AI as an independent channel of negotiation, monitoring, and representation, not as a unilateral tool of the company.
A legitimate labor AI system would need to obey minimum principles:
transparency of criteria;
external auditing;
shared governance;
data protection;
explainability of recommendations;
human review in sensitive decisions;
direct voting by workers;
historical records of negotiations;
open simulations;
prevention of capture by the company, the state, or political parties.
In other words: AI should replace inefficient union intermediation, not abolish worker defense.
The Professional Unionist Facing Automation
There is an almost poetic historical irony in this process.
For decades, unions denounced automation as a threat to workers. Now automation threatens to automate part of unionism itself.
And why not?
If bank tellers were automated, if attendants were replaced by apps, if industrial operators began working with robots, if teachers use digital platforms, if doctors use diagnostic AI, why should the union leader be the last species protected from technological selection?
Perhaps it is time to ask: how many union processes are truly work of representation, and how many are merely bureaucratic liturgy?
Calling an assembly: automatable.
Collecting demands: automatable.
Organizing an agenda: automatable.
Comparing salaries: automatable.
Analyzing an employer proposal: automatable.
Simulating economic impact: automatable.
Translating legal jargon: automatable.
Monitoring compliance with clauses: partially automatable.
Detecting algorithmic abuse: automatable with auditing.
Delivering a 47-minute fiery speech blaming neoliberalism, imperialism, technology, and the “correlation of forces”: unfortunately, also automatable — but no one should want to preserve that function.
What remains for the human union? Perhaps what should always have been its noble function: listening, solidarity, legitimate organization, defense in extreme cases, protection of the vulnerable, and democratic control of the tools.
But that would require humility. And humility has never exactly been the native software of bureaucracies.
AI Negotiation: Fiction? Less and Less.
Recent research on autonomous negotiation agents shows that AI models can already conduct complex negotiations at scale, test strategies, seek agreements, and adapt behavior. Studies on AI negotiation coaching also indicate that systems can help individuals prepare better, reduce fear, and rehearse arguments, although they still have limitations in clarity, structure, and trust.
This suggests a realistic path: not a sovereign AI deciding everything, but an ecosystem of assisted negotiation.
The worker could use AI to understand rights, compare proposals, and prepare arguments. The company could use AI to estimate financial and operational impacts. An independent body could use AI to mediate scenarios, identify points of consensus, and suggest alternatives. The regulator could use AI to detect patterns of abuse. The worker would vote directly. The final decision would remain human, but far more informed.
It would be a kind of “algorithmic union” — not in the sense of an authoritarian machine, but as an intelligent infrastructure for representation.
Less sound truck. More dashboard.
Less slogan. More evidence.
Fewer captured assemblies. More direct consultation.
Fewer eternal leaders. More worker at the center.
The Obvious Objection: What About Algorithmic Capture?
The strongest criticism of this proposal is legitimate: who controls the AI?
If the company controls it, the worker loses.
If the government controls it, the worker may become a political statistic.
If the union controls it, AI may become merely an automated pamphlet.
If a big tech company controls it, labor conflict becomes a product.
Therefore, the issue is not only technical. It is institutional.
A truly useful labor AI system would need plural governance, auditable code when possible, models evaluated by third parties, robust data protection, the possibility of contestation, and transparency mechanisms. Moreover, sensitive final decisions — dismissal, deactivation, punishment, pay reduction, schedule changes — should not be made exclusively by automated systems.
The European Union is already moving in this direction by requiring human supervision in relevant decisions involving platform work. The OECD recommends consulting workers to reduce risks and increase acceptance of these tools. The ILO emphasizes that social dialogue can guide AI to complement workers, rather than simply replace them.
These points do not weaken the argument of this article. On the contrary: they show that the real world has already understood that AI and work have become inseparable.
The question is whether workers will be represented by data, simulations, and transparent mechanisms — or by leaders who still believe technology is a threat because it allows the base to speak without asking permission.
The New Labor Negotiation
The traditional model of labor negotiation is slow, episodic, and centralized. The future model tends to be continuous, personalized, and data-driven.
Instead of waiting for the annual bargaining date, workers could monitor indicators of wages, inflation, productivity, profits, turnover, and satisfaction in real time. Instead of depending on a union board to know whether a proposal is good, they could compare scenarios automatically. Instead of voting in empty assemblies, they could vote digitally with secure authentication. Instead of receiving ideological pamphlets, they could receive clear reports.
Negotiation would stop being an event and become a system.
Companies would also benefit. Strikes are expensive. Prolonged conflicts destroy trust. Poorly communicated decisions create resentment. An AI mediator could help build more acceptable proposals, identify bottlenecks before they explode, and create solutions in which productivity gains are shared more rationally.
Naturally, this bothers those who live from the explosion.
There are mediation professionals who need conflict to remain opaque. After all, a conflict that is too clear can be solved; and a conflict that is too solved reduces the demand for specialists in administering eternal conflicts.
There are bureaucracies that prefer workers to remain uninformed. An informed worker votes better, questions more, compares promises with results, and, unforgivable sin, obeys less. Any resemblance to certain habits of contemporary Brazilian democracy is purely coincidental. Or perhaps not coincidental at all.
There are leaders who claim to represent the base, but panic at the simple possibility of asking the base directly what it actually wants. Because one thing is to speak “in the name of the people.” Another, far more dangerous thing, is to let the people answer.
AI threatens exactly this theater.
The End of the Union Monopoly
The future probably will not be “without unions.” It will be worse for traditional unions: it will be a future in which they must compete.
Compete with direct consultation apps.
Compete with automated legal guidance platforms.
Compete with algorithmic auditing systems.
Compete with decentralized worker communities.
Compete with AI capable of explaining contracts better than many lawyers.
Compete with economic simulations more precise than grandstanding speeches.
Compete, finally, for the worker’s real trust.
And that is the point: representation should be earned, not presumed.
If a union delivers value, let it survive. If it protects workers, let it flourish. If it uses AI to expand transparency, excellent. If it organizes data, listens to the base, and negotiates intelligently, it will have a future.
But if it depends on compulsion, opacity, ideology, manipulated assemblies, fear of technology, and nineteenth-century rhetoric, perhaps its destiny is the museum — next to the fax machine, the mimeograph, and minutes read aloud to a room with 17 people and three banners.
Conclusion: Less Union, More Intelligence
Artificial intelligence will not eliminate the conflict between capital and labor. That tension is structural. Companies will continue to seek efficiency. Workers will continue to seek safety, income, and dignity. What can change is the channel of negotiation.
AI can make that channel faster, more transparent, more personalized, more democratic, and less dependent on professional intermediaries. It can help workers negotiate better, companies propose better, governments monitor better, and society reduce unnecessary conflicts.
But this will only happen if AI is designed as an infrastructure of balance, not as a digital whip. The desirable future is not the employer’s algorithm replacing the union. It is the worker using AI to depend less on both the employer and the professional unionist.
Deep down, perhaps this is the true revolution: not the revolution promised in endless speeches, but the silent revolution of a tool that gives workers something they have always been denied by paternalistic employers and messianic unionists alike: clear information, power of choice, and direct voice.
If I were a professional unionist, I would be worried.
The tools are already here. Workers are already connected. The data already exist. AI is learning to negotiate, explain, simulate, audit, and represent interests. It is only a matter of time before many realize they do not need a lifetime intermediary to translate their own needs.
And when that day comes, perhaps professional unionists will discover, moved to tears, the human drama of technological obsolescence.
Maybe then they will organize an assembly to discuss the issue.
With in-person voting, naturally.
Because tradition is tradition.
#ArtificialIntelligence #AI #FutureOfWork #DigitalWork #Unions #LaborNegotiation #AlgorithmicManagement #AppDrivers #Uber #GigEconomy #DigitalPlatforms #Technology #Innovation #AIAtWork #UnionBureaucracy
References
Directive (EU) 2024/2831. (2024). Improving working conditions in platform work. Official Journal of the European Union.
Doellgast, V., Appalla, S., Ginzburg, D., Kim, J., & Thian, W. L. (2024). Global case studies of social dialogue on AI and algorithmic management (ILO Working Paper No. 144). International Labour Organization.
Milanez, A., Lemmens, A., & Ruggiu, C. (2025). Algorithmic management in the workplace: New evidence from an OECD employer survey (OECD Artificial Intelligence Papers No. 31). OECD Publishing.
Milanez, A. (2025). Exploring win-win outcomes of algorithmic management: Lessons from a laboratory experiment on worker consultation (OECD Artificial Intelligence Papers No. 43). OECD Publishing.
Rani, U., Pesole, A., & González Vázquez, I. (2024). Algorithmic management practices in regular workplaces: Case studies in logistics and healthcare. European Commission, Joint Research Centre, and International Labour Organization.
SAG-AFTRA. (2024). Interactive Media Agreement. Screen Actors Guild-American Federation of Television and Radio Artists.
Stanford Institute for Human-Centered Artificial Intelligence. (2024). Artificial Intelligence Index Report 2024. Stanford University.
Writers Guild of America. (2023). Artificial Intelligence — 2023 MBA protections.
NewsGuild of New York. (2026, May 27). NewsGuild of NY, Tech Guild take legal action against The New York Times. NewsGuild of New York.
Roth, E. (2026, May 28). The AI fight brewing inside The New York Times. The Verge.
Read also: The Ultimatum Game: Where Rational Choice Meets Fairness, Emotion, and Artificial Intelligence
Copyright 2026 AI-Talks.org
