Protecting People in the Age of AI:
A Shared Commitment to Dignity, Opportunity, and Security
The defining challenge of the AI era is whether we can create progress without treating people as disposable.
During the May 2026 graduation season, a striking response echoed across several American campuses: graduates loudly booed commencement speakers who praised artificial intelligence.
At the University of Arizona, former Google CEO Eric Schmidt was met with jeers while discussing AI’s role in shaping the future. At the University of Central Florida, Gloria Caulfield, vice president of strategic alliances at Tavistock Development Company, drew boos after calling AI “the next industrial revolution.” At Middle Tennessee State University, graduates responded similarly when music executive Scott Borchetta, founder, chairman and CEO of Big Machine Records, told them that “AI is rewriting production.”
It would be easy to dismiss these reactions as resistance to change. But to understand them, we must begin by listening.
The students were not necessarily rejecting technology. They were rejecting a story about progress that seemed to have no place for them in it.
They had gathered to celebrate years of study, sacrifice and hope. Many had worked while attending school, taken on debt, relied on family support or overcome significant barriers simply to reach graduation day. Now, standing at the threshold of their working lives, they were looking for reassurance that their effort still mattered—and that a meaningful future remained open to them.
Instead, repeated calls to embrace AI could sound less like encouragement and more like a warning.
Graduates are entering a difficult entry-level job market while companies increasingly describe AI efficiency in terms that suggest fewer opportunities for people beginning their careers. They are weary of being told to welcome a technology they fear may weaken critical thinking, devalue creativity and narrow their economic security.
To a business leader, talk of an AI-powered industrial revolution may sound optimistic.
To a graduate searching for a first opportunity, it can sound very different:
The work you trained to do may be automated before anyone gives you the chance to prove yourself.
That is why the students pushed back. Their boos were not merely hostility toward innovation. They carried fear, grief and moral clarity. The graduates were asking whether those celebrating disruption had considered who would absorb its cost.
Would a degree still open a door? Would creative work still be valued? Would entry-level roles still exist? Had the institutions that urged them to prepare for the future prepared a future that included them?
Their anxiety is justified.
Recent research does not establish that AI is causing economy-wide mass unemployment. It does, however, show that AI can perform or accelerate many of the tasks through which people earn a living—and through which beginners become experienced professionals.
Three randomized field experiments involving 4,867 software developers at Microsoft, Accenture, and an anonymous Fortune 100 company found that, when the results were combined, developers given access to an AI coding assistant completed 26% more tasks on average. That finding demonstrates genuine productive power.
But productivity is not the same as assurance. The study measured completed tasks—not whether the resulting software was more accurate, secure, maintainable, or aligned with business intent. From a governance perspective, faster output should be evaluated alongside quality, defects, rework, security, traceability, and accountable Human review.
It also raises a deeply human question that no productivity statistic can answer:
When less labor is needed to produce the same output, who receives the benefit—and who carries the risk?
That question should concern every community leader, legislator, technology executive, employer and AI investor.
The defining challenge of the AI era is not simply whether we can build more capable systems.
It is whether we can use them to create progress without treating people as disposable.
A job is more than a collection of tasks
Economic discussions often reduce work to labor hours, productivity, headcount and cost.
But a job is also rent paid, medicine purchased, a child cared for and a future planned.
It can be a source of identity, belonging, confidence and contribution.
Losing work—or never being given a first opportunity—can mean more than losing income. It can mean losing stability, social connection and the belief that one has a valued place in the world.
This is especially important when discussing entry-level work. Junior tasks may look routine to an experienced leader or easily automatable to an engineer. Yet those tasks are often where people learn how a profession actually works. A first draft teaches judgment after it is reviewed. A basic coding assignment becomes a lesson in reliability after it fails. A junior analyst learns what matters by watching how an experienced colleague challenges assumptions. A new customer-support worker develops empathy by encountering people whose needs do not fit the script.
If AI removes these tasks without replacing the learning pathways around them, society may not only eliminate jobs. It may remove the first rungs of the ladder while continuing to demand that people somehow arrive at the top.
Employers should therefore redesign entry-level roles, not erase them.
Young workers can use AI while receiving supervised practice in verification, communication, ethical judgment and responsibility.
Paid apprenticeships, residencies, internships and rotational programs should be treated as investments in the future capacity of a profession—not as charitable extras to be cut when automation improves.
Innovation without empathy is an incomplete achievement
AI can help people write, code, analyze information, serve customers, discover patterns and create new products. Those capabilities can be genuinely valuable. They may reduce drudgery, expand access to expertise and allow smaller organizations to accomplish more. We should not deny those possibilities.
But possibility is not destiny. Technology does not decide how productivity gains are distributed. People, companies, markets and governments do.
A company can use AI to support workers or to intensify surveillance.
It can reduce hours while preserving pay or eliminate positions while concentrating the savings.
It can retrain people during paid working time or tell them to reinvent themselves alone.
It can preserve human review for consequential decisions or hide accountability behind an algorithm.
It can maintain junior opportunities or enjoy the short-term savings of eliminating them while leaving the next generation without a path into the profession.
Every one of these is a human choice presented as a business decision.
The language we use matters as well. Workers are not “legacy costs.” A person whose tasks can be automated has not become obsolete.
Experience does not lose its worth because a model can generate an answer quickly.
People carry context, relationships, ethical responsibility, practical wisdom and knowledge of consequences.
They also carry obligations—to families, communities and themselves—that do not disappear when an efficiency target is met.
Leaders should speak about AI transformation with humility.
Before celebrating a reduction in labor hours, they should be able to say what will happen to the people whose hours are no longer needed.
Before announcing an AI-first strategy, they should explain how workers will participate in its design, how harms will be measured and how gains will be shared.
Before calling displacement inevitable, they should acknowledge the decisions that made it so.
Those who benefit most from AI should bear meaningful responsibility for managing its disruption.
For employers, that responsibility begins before deployment.
Any AI system that may affect hiring, staffing, pay, workload, promotion, surveillance or termination should undergo a workforce-impact assessment. The people affected should receive notice early enough to influence the decision. Workers and their representatives should have a voice in workflow design, safety testing and performance measurement. No person should be hired, disciplined or dismissed solely through an automated process without a named human decision-maker and an accessible appeal.
When roles change, redeployment should come before displacement. Employers should provide paid learning time, salary and benefit continuity during transitions, and first consideration for newly created work. Experienced employees should be invited to shape safer processes and train others, not treated as resistant simply because they understand what can go wrong.
When displacement truly cannot be avoided, dignity requires more than legal-minimum severance. People need time, income support, healthcare continuity, credible training, independent career assistance and a fair opportunity to move into comparable work. Contractors and lower-wage workers deserve protection too; employment classification should not determine whose hardship counts.
And when AI creates measurable value, workers should share in it. That may take the form of higher wages, bonuses, profit sharing, employee ownership, reduced hours without reduced pay or adequately funded transition programs.
Productivity should improve human life—not merely make human labor easier to remove.
For AI developers, cloud providers and investors, responsibility extends beyond the companies they directly employ. Their products and capital shape incentives throughout the economy. They can require workforce-impact reporting, support independent audits, fund transition programs and make responsible labor practices part of investment and procurement decisions.
They can ask not only, “Can this company scale?” but also, “Who becomes more vulnerable if it does, and what protections are built into the model?”
A return on investment should not depend on an unpriced transfer of risk to workers, families and communities.
Public leadership must protect agency, not only encourage adaptation
Governments cannot outsource the social contract to corporate goodwill. Responsible companies should act, but voluntary promises will never be enough when competitors can gain an advantage by externalizing the costs of displacement.
Legislators should create a public floor: meaningful notice before technology-driven workforce reductions; human review of consequential employment decisions; protection from discriminatory and intrusive workplace AI; modern unemployment and wage insurance; healthcare continuity; paid, demand-linked training; portable benefits; stronger worker-data rights; and transparent reporting of AI-related hiring reductions, job redesign and displacement.
Communities also need support when disruption is concentrated geographically. A laid-off worker may retrain, but an entire region cannot simply “learn AI” and move away.
Community colleges, public workforce systems, unions, local employers and civic organizations should be funded to build pathways tied to real jobs.
Where AI-driven restructuring weakens a local economy, investment should support small businesses, infrastructure, caregiving, skilled trades, public-interest technology and other forms of durable community capacity.
Policy should preserve people’s agency. Assistance should not force someone to accept unsafe, unstable or sharply lower-paid work merely to remain eligible for support. Training should not be measured by enrollment numbers while graduates remain unemployed.
A humane system asks whether people regained income, stability, choice and a credible future.
The people most affected must have a seat at the table
Nothing about people without people.
Students and recent graduates should help design early-career programs. Workers should participate in decisions about systems that measure or manage them. Artists and creators should be included in rules governing the use of their work. People with disabilities, caregivers, contractors, older workers and communities historically excluded from technological prosperity should be present before policies are finalized—not invited afterward to react.
This is not a barrier to innovation.
It is how innovation becomes more legitimate, informed and durable.
People closest to a workflow often know where an automated system will fail, where hidden labor is being shifted and where an efficiency claim ignores a human consequence. Their knowledge is not resistance. It is evidence.
A different measure of progress
We need a broader definition of success for the age of AI.
Progress cannot be measured only by model capability, adoption rates, productivity or market valuation.
It must also be measured by whether people can still enter a profession, support a family, exercise judgment, contest an automated decision and share in the wealth their work and knowledge helped create.
A successful AI transition would mean that productivity rises while dignity remains intact.
Workers would have meaningful influence over systems that affect them.
Students would graduate into redesigned opportunities rather than disappearing career ladders.
People who lose roles would receive real support rather than advice to become more adaptable.
Communities would gain new capacity instead of inheriting the cost of private efficiency.
Leaders would be accountable not only for what their systems can do, but for what happens to people when those systems are deployed.
The graduates who booed in May 2026 were asking adults in positions of power to see the future from where they stood.
Their message was not that technology must stop. It was that progress which demands human sacrifice without consent, protection or shared benefit is not progress enough.
We should hear them.
The age of AI will test more than our intelligence. It will test our conscience: whether we treat efficiency as the highest good, whether we confuse what is possible with what is just and whether those with power will protect people who have less of it.
We can choose a future in which AI expands human capability while institutions preserve human security.
We can choose to make adaptation a shared responsibility rather than a private burden.
We can choose to value care, creativity, experience and judgment even when they are difficult to quantify.
At FlyWheel Angel OS, that choice is expressed in our Human-Centered AI Commitment:
FlyWheel Angel OS is grounded in the conviction that AI should remain under accountable Human judgment and be guided by empathy and care. We believe no innovation is complete until the people it affects have voice, protection, and a fair share of the value it creates. We choose AI that expands human capability while preserving human agency, dignity, and security.
We offer this commitment not as a finished answer or a standard owned by one company, but as an open invitation—for employers, developers, investors, and institutions to make their own public, measurable commitments and remain accountable to the people most affected by their decisions.
That is how we protect people in the age of AI: by innovating thoughtfully, challenging disruption that harms people, and making sure technology remains in service to human lives.
Frequently Asked Questions
Many of us are asking the same urgent questions about AI and the future of work. The FAQs below bring together those shared questions—not as ideas unique to this article, but as concerns already being voiced in workplaces, classrooms, families and public debate. I researched the answers using the sources listed below to offer readers an informed view and help distinguish what is known from what remains uncertain.
Fear of AI-driven job loss is justified when we consider exposed tasks, particular occupations and vulnerable career stages. AI is already increasing productivity in customer support, professional writing, consulting and software development, while emerging evidence points to weaker hiring outcomes for some young workers in exposed occupations.
At the same time, evidence through August 2026 does not show broad, AI-caused mass unemployment. Measures of exposure alone also cannot tell us whether employment will ultimately rise or fall.
The central danger is not that all human work will vanish on a known timetable. It is that organizations may automate entry-level tasks, consolidate routine cognitive work, raise output expectations and weaken bargaining power faster than societies strengthen training, income protection, worker voice and access to ownership.
Whether AI primarily substitutes for people or expands what people can do will be shaped by the choices we make—in product design, employment practices, labor institutions, regulation and the distribution of productivity gains.
Q: Are humans’ fears that AI will take their jobs validated?
Answer: Partly. AI is already automating or accelerating tasks that people are paid to perform, especially digital, language-based, document-heavy and codifiable work. That creates real risks of reduced hiring, job redesign and displacement in particular occupations. But the evidence reviewed for this article through August 2026 does not show broad, economy-wide mass unemployment caused by AI. The International Labour Organization (ILO) concludes that job transformation is the most likely general effect because most occupations contain a mix of exposed and unexposed tasks; the OECD likewise finds high exposure without an aggregate employment decline attributable to AI so far.[1][2]
The most important distinctions are:
AI exposure: AI can perform or materially accelerate some tasks in an occupation. Exposure does not establish that a person will lose a job.
Task substitution: AI performs work that a person previously performed.
AI complementarity: AI helps a person perform work faster, better or at a larger scale.
Job displacement: a worker loses employment, income or opportunities because labor demand falls.
Job transformation: the occupation remains, but its task mix, skill requirements, staffing level or career path changes.
Observed exposure remains much lower than theoretical capability. Anthropic’s company-authored, usage-based analysis found that real-world AI coverage was still a fraction of what language models could theoretically perform and found no systematic rise in unemployment among workers in the most exposed occupations. It did, however, find suggestive evidence of slower hiring into exposed occupations for workers aged 22–25.[3]
The evidence therefore supports concern about specific workers, tasks and career pathways, not a conclusion that work as a whole is disappearing.
Q: What jobs can AI take away from humans?
Answer: AI does not usually replace an occupation all at once. It first changes or removes tasks. Jobs face greater substitution pressure when a large share of their work is digital, repetitive or codifiable; can be completed and evaluated through software; and does not require sustained physical action, responsibility for high-consequence decisions or deep interpersonal trust. ILO’s refined index evaluates nearly 30,000 tasks and finds clerical occupations remain the most exposed, while exposure has also increased in highly digitized professional and technical work.[4]
1. Clerical and administrative work
The clearest high-exposure category includes data entry, typing, bookkeeping, payroll, scheduling, document processing, records administration and general office support. AI can extract data, classify and route documents, draft routine correspondence, reconcile information and update software systems. These capabilities can reduce the amount of human time required even when a person remains responsible for exceptions and quality control.[5][1]
Likely effects include fewer purely transactional positions, consolidation of several administrative workflows into broader operations roles, and greater emphasis on exception handling, coordination, stakeholder communication and accountability.
2. Customer support and call-center work
Routine chat, email and scripted service interactions are highly exposed. A large field study of 5,172 customer-support agents found that access to a generative-AI assistant increased issues resolved per hour by about 15% on average, with the largest gains among less-experienced workers. That is direct evidence of augmentation and productivity improvement—not proof that all agents will be replaced—but it also demonstrates that the same workload may eventually require fewer labor hours if demand does not expand.[6]
Human work remains especially important for complex exceptions, emotionally sensitive cases, fraud, regulated decisions, safety issues, high-value relationships and cases in which someone must be accountable for the outcome.
3. Routine writing, translation and content production
Generative AI can draft, summarize, edit and translate text and can produce images, audio, video and presentation content. In a randomized experiment involving 453 college-educated professionals completing occupation-specific writing tasks, ChatGPT reduced completion time by 40% and increased assessed quality by 18%. These results apply to bounded writing assignments; they do not establish equivalent performance in investigative reporting, original creative direction, sensitive communications or work requiring verified facts and legal accountability.[7]
The most exposed work includes standardized product descriptions, routine marketing variants, first drafts, basic summaries, transcription and straightforward translation. More resilient responsibilities include editorial judgment, original reporting, brand strategy, audience insight, source verification, cultural interpretation, creative direction and accountability for published claims.
4. Software development and technical work
AI can generate code, tests and documentation; explain code; translate between languages; and assist with debugging. Randomized field experiments involving 4,867 developers at Microsoft, Accenture and a Fortune 100 company found a combined 26% increase in completed tasks among developers given access to a coding assistant, with greater gains and adoption among less-experienced developers. This supports substantial task-level productivity effects, not the disappearance of software engineering.[8]
Boilerplate code, routine tests, documentation and well-specified fixes are more exposed than architecture, security, reliability, complex debugging, requirements discovery and cross-system tradeoffs. The labor-market concern is that employers may need fewer junior labor hours for routine work, potentially narrowing the entry-level pathway through which developers acquire experience.
5. Legal, financial, accounting and professional-services support
Document review, research summaries, due-diligence support, financial drafting, invoice classification, compliance documentation and standardized analysis are exposed because they are text- and data-intensive. Experiments with consultants show that AI can improve speed and quality on tasks within the technology’s capability frontier, but performance can fall when users rely on AI for tasks outside that frontier. Domain expertise and verification therefore become more—not less—important in consequential professional work.[9][10]
Negotiation, fiduciary and legal duties, regulatory interpretation, client trust, ethical judgment and final accountability remain human responsibilities even when AI prepares part of the work.
6. Research and analytical support
Literature scans, evidence summaries, first-pass comparisons, spreadsheet assistance, market research and report formatting are exposed. AI can accelerate these components, but reliable research still requires sound question formulation, source selection, provenance, methodological judgment, contradiction checking and interpretation. Because fluent output can contain fabricated or weakly supported claims, consequential research requires human validation and traceable sources.[11]
7. Jobs less exposed to full replacement by current AI
Current generative AI is less capable of replacing jobs centered on physical manipulation in unpredictable settings, sustained human care, relationship-based trust, real-time responsibility or embodied presence. Examples include electricians, plumbers, HVAC technicians, many construction and maintenance roles, nurses and care workers, physical therapists, emergency responders, early-childhood educators and high-touch service roles. These occupations may still use AI for scheduling, diagnosis support, documentation or training, but automating the whole job would also require capable and economically viable robotics, safe deployment, regulatory acceptance and public trust.[2][3]
Q: How far are those fears from becoming reality?
Answer: The effects are already real at the task and workflow level, but the pace of whole-job displacement remains uncertain. The sources reviewed do not establish a reliable timetable in which broad categories of jobs disappear within two, five or ten years. Adoption depends on model capability, reliability, cost, regulation, data access, integration with business systems, organizational redesign, customer acceptance and the demand created by lower costs and new products.[12]
What the evidence shows now
Substantial productivity effects exist in bounded tasks. Controlled and field studies document faster or higher-quality work in professional writing, customer support, consulting and software development.[7][6][8]
The reviewed studies do not show widespread displacement. OECD analysis found no negative aggregate employment outcome associated with AI exposure in its 2012–2022 data. Anthropic found no systematic unemployment increase among highly exposed workers, and Stanford’s updated working paper found no evidence of widespread economy-wide displacement in its payroll sample. These findings do not prove that AI has no employment effects or predict what will happen as adoption expands.[2][3][13]
Entry-level effects are an important early warning. Stanford researchers using ADP payroll records through June 2026 found that employment of workers aged 22–25 in AI-exposed occupations stood 19% below where it would have been had it kept pace with less-exposed peers. The difference operated mainly through reduced hiring rather than increased separations. The authors explicitly describe the result as descriptive—not a causal estimate—and note preexisting trends, education-related confounding and sample-representativeness limits.[13]
The way AI is used matters. The same Stanford analysis found declines concentrated where observed AI use substituted for human tasks; employment was flat or rising where AI primarily complemented workers. Anthropic similarly finds limited aggregate effects but tentative evidence of weaker job-finding rates for young workers entering highly exposed occupations.[3][13]
Exposure does not imply inevitable decline. OECD analysis of 2012–2022 data found employment growth was stronger, not weaker, in more AI-exposed occupations, illustrating why capability measures alone cannot predict labor demand. That historical relationship may change as generative AI adoption deepens, so it should not be treated as a guarantee.[2]
The most defensible near-term concern is therefore not immediate mass unemployment. It is a combination of slower hiring in selected occupations, elimination of some routine tasks, role consolidation, higher output expectations, reduced freelance demand in easily standardized work and weakening entry-level career ladders. A labor-market shock could appear in hiring and occupational mobility before it appears in aggregate unemployment.
Q: If AI reduces demand for some jobs, how can people continue to earn a living?
Answer: There is no universally “AI-proof” occupation, and “learn to code” is not a sufficient answer because coding itself is highly exposed. The evidence favors a strategy of combining AI fluency, domain expertise, judgment, interpersonal capability and responsibility for outcomes. Productivity studies repeatedly find that humans and AI perform best when the task fits the technology and users can recognize, verify and correct its limitations.[10]
1. Become an AI-augmented practitioner in an existing field
Many workers can remain in their field while moving from routine production toward orchestration, quality control, exceptions and accountable decision-making. Examples include:
| Exposed work | More resilient responsibilities |
| Routine administration | Operations coordination, stakeholder management and exception handling |
| Scripted support | Escalations, retention, complex troubleshooting and customer success |
| First-pass analysis | Domain-specific interpretation, source validation and decision support |
| Routine copy production | Editorial judgment, brand strategy, original reporting and creative direction |
| Standard legal or finance support | Compliance operations, controls, client judgment and accountable review |
| Boilerplate development | Product engineering, systems integration, security and reliability |
| Data entry | Data quality, provenance, workflow assurance and exception resolution |
These are transition directions, not guarantees. Their resilience depends on actual regional demand, credential requirements and whether employers redesign jobs to complement workers rather than simply reduce headcount.
2. Build skills that complement AI
Useful capabilities include domain expertise, problem framing, critical evaluation, data literacy, communication, negotiation, ethical reasoning, workflow design, regulatory knowledge and the ability to take responsibility for a decision. Experimental evidence also warns that AI can lower performance on tasks outside its capability frontier, which makes verification and knowing when not to use AI essential skills.[9]
3. Move toward care, trust and the physical world where appropriate
Healthcare, caregiving, teaching, skilled trades, field service, maintenance, relationship management and safety work are generally less exposed to full replacement by language models. These pathways are not automatically accessible or well paid; they may require licensing, apprenticeships, physical capacity, public investment and improvements in job quality. They are options—not a universal prescription.
4. Work in AI governance, assurance and oversight
As organizations deploy AI in employment, finance, healthcare, education, public benefits and other consequential settings, they need risk management, testing, documentation, monitoring, incident response, compliance and human-oversight capabilities. NIST’s AI Risk Management Framework provides a voluntary structure for managing risks to people, organizations and society. The EU AI Act treats employment, education, credit and certain essential services as high-risk uses and requires controls including risk management, data governance, logging, documentation, accuracy and human oversight.[11][14]
Potential work includes AI risk analysis, model and workflow evaluation, compliance, audit support, data provenance, incident investigation, algorithmic-impact assessment, red teaming and human-in-the-loop operations. The number and quality of these jobs will depend on regulation and adoption and should not be assumed to offset every displaced position.
5. Use AI in entrepreneurship cautiously
Generative AI can lower some entry barriers by reducing the time and cost of drafting, research, marketing, coding and routine business administration. The OECD finds promising experimental evidence for productivity, innovation and entrepreneurship, but emphasizes that long-term firm-level research remains scarce and that outcomes depend heavily on task fit, expertise and verification.[10]
Entrepreneurship can create options for some people, but it cannot serve as the sole social response to displacement. Access to capital, healthcare, time, networks, customers and risk tolerance is unequal, and AI-generated advice can harm performance when users cannot evaluate it effectively.
Q: What intentional, meaningful and humane steps should AI-benefiting companies take?
Answer: Companies that capture revenue, cost savings or productivity gains from AI should accept a corresponding duty to protect the people who helped create that value. The goal should not be to prevent every job from changing. It should be to ensure that transformation is transparent, workers have a meaningful voice, people receive a fair opportunity to adapt, and those who are displaced do not bear the full cost of a transition that benefits employers and investors.
These commitments should apply to employees, contractors and contingent workers. They should also distinguish between experienced workers—who may face income loss, skill devaluation or late-career displacement—and students and recent graduates, who may lose the entry-level opportunities through which professional judgment is normally developed.
Actions each company can take
Adopt a human-impact standard before deploying AI.
Require an AI workforce-impact assessment before any deployment that could affect staffing, hiring, hours, compensation, workload, surveillance or promotion.
Document which tasks will be automated, which will be augmented, who may be harmed, what alternatives were considered and who is accountable for the decision.
Test whether claimed savings depend on transferring hidden work, risk or uncompensated review duties to employees or customers.
Reassess impacts after deployment using actual outcomes rather than vendor claims.
Give workers meaningful notice and participation.
Inform affected workers early enough to influence the plan—not after the system has been purchased and layoffs have been decided.
Consult employees, worker councils, unions and affected contractors about workflow design, safety, performance measurement and staffing.
Provide a named human decision-maker and a documented appeal process when AI contributes to hiring, evaluation, discipline, scheduling, pay or termination.
Prohibit retaliation against workers who report unsafe, discriminatory or misleading AI practices.
Make redeployment the default before displacement.
Conduct a skills inventory and offer qualified workers first consideration for newly created or redesigned roles.
Create paid transition pathways into operations, customer success, quality assurance, AI evaluation, security, compliance, data stewardship and other areas of demand.
Allow sufficient paid learning time during working hours. Do not require people to retrain at night or at personal expense while the company captures the productivity benefit.
Continue salary and benefits during a defined transition period and publish placement outcomes.
Share productivity gains with the people who produce them.
Establish a transparent formula that directs part of verified AI-related savings or revenue into wage growth, bonuses, profit sharing, employee ownership, reduced hours without reduced pay, and a worker-transition fund.
Avoid using AI solely to raise output quotas or eliminate recovery time. Productivity gains should improve job quality as well as margins.
When the same output can be produced with fewer labor hours, consider shorter workweeks, voluntary reduced schedules and work sharing before layoffs.
Provide a humane displacement floor when job loss cannot be avoided.
Offer severance that reflects tenure, compensation and the difficulty of reemployment—not merely the legal minimum.
Continue healthcare or provide an equivalent benefit bridge; maintain access to counseling, financial planning, immigration support where relevant and independent career services.
Fund recognized training, licensing, apprenticeships or education selected with the worker, rather than restricting support to a narrow vendor course.
Provide advance notice, paid job-search time, references and access to internal vacancies.
For lower-wage workers and contractors, create a minimum transition payment so protection does not depend on employment classification or bargaining power.
Protect experienced workers from forced obsolescence.
Do not treat higher compensation or long tenure as evidence that a worker is less adaptable.
Pair experienced practitioners with technical teams to preserve institutional knowledge and translate it into safer workflows, controls and training.
Offer phased transitions, reduced schedules, mentoring roles and bridge-to-retirement options without coercion.
Evaluate redesigned roles for age and disability discrimination, including whether new AI-based productivity targets create indirect exclusion.
Preserve entry-level career ladders for students and recent graduates.
Maintain paid internships, apprenticeships, residencies and junior roles even when AI can perform part of the routine work.
Redesign—not erase—entry-level jobs: graduates can use AI for first-pass tasks while receiving supervised practice in verification, client interaction, judgment, ethics and accountability.
Set a minimum ratio of paid early-career positions or training hours relative to senior and AI-enabled roles where junior pipelines are essential to the profession.
Use skills-based hiring, publish salary ranges and remove unnecessary experience requirements that block graduates from jobs formerly used to gain that experience.
Partner with colleges and community colleges to align curricula with real roles, but pay for work-based learning and do not shift the full cost of employer training onto students.
Track early-career applications, interviews, hiring, retention, wages and promotion separately so a weakening pathway is visible before it becomes a profession-wide shortage.
Set limits on high-risk and dehumanizing uses.
Do not allow fully automated termination, discipline or other consequential employment decisions.
Ban emotion recognition and covert behavioral inference in employment where validity, necessity and proportionality cannot be demonstrated.
Minimize worker data collection, set deletion periods, restrict secondary use and give workers access to consequential data about them.
Require independent testing for discrimination, security, accuracy and foreseeable misuse, consistent with risk-management principles such as the NIST AI Risk Management Framework.[[11]](https://www.nist.gov/itl/ai-risk-management-framework)
Report outcomes, not only intentions.
Publish the number of roles created, eliminated, left unfilled and materially redesigned because of AI; disclose effects by occupation, level, location, employment status and relevant demographic groups where lawful and privacy-preserving.
Report savings, transition spending, redeployment rates, post-transition wages, training completion, involuntary separations and use of contractors.
Commission independent audits and allow worker representatives to review methodology.
Tie executive compensation to safe deployment, successful redeployment, job quality and equitable distribution—not only adoption speed or headcount savings.
Q: What Actions a collective of AI-benefiting companies can take?
Answer: Individual programs are vulnerable to competitive pressure: a company that invests in workers may be undercut by one that externalizes transition costs. Employers, AI developers, cloud providers, investors and major deployers should therefore create shared institutions and enforceable commitments.
Create a pooled AI transition and opportunity fund.
Finance it through member dues linked to AI revenue, compute use, productivity savings or the scale of AI-related workforce change.
Use the fund for portable income support, healthcare continuity, accredited training, apprenticeships, relocation assistance and community recovery.
Give workers, educators and affected communities voting power over fund governance; do not leave allocation solely to the companies creating the disruption.
Establish portable benefits and learning accounts.
Make protections follow the person across employers and employment classifications.
Fund accounts that can pay for credentials, licensing, tools, caregiving, transportation and other barriers to participation—not just online courses.
Design benefits so contractors, freelancers and workers at smaller suppliers are not excluded.
Build a cross-company talent exchange.
Give displaced workers verified skill records, priority access to open roles, interoperable credentials and paid bridge training across participating companies.
Require receiving employers to recognize demonstrated skills and avoid resetting experienced workers to entry-level pay solely because a job title changed.
Establish rapid placement partnerships with public workforce systems, unions, professional associations and educational institutions.
Guarantee an early-career opportunity pipeline.
Jointly fund paid apprenticeships, fellowships and first-job programs in fields where AI is reducing junior tasks.
Rotate participants across companies so they can acquire broad experience without any one employer carrying the full cost.
Reserve meaningful participation for first-generation students, community-college graduates and people from regions or groups most exposed to lost opportunities.
Measure success through conversion to durable, fairly paid employment—not the number of people enrolled.
Adopt common workforce-impact disclosures.
Use shared definitions for AI-related hiring reductions, displacement, task transformation, wage effects and productivity gains.
Support confidential reporting to public statistical agencies and publish comparable aggregate data.
Fund independent labor-market research with protections against sponsor control, enabling early identification of occupational and regional harm.
Set sector-wide minimum standards.
Agree on minimum notice, severance, healthcare continuation, paid retraining, human review and worker consultation.
Extend standards through procurement contracts so suppliers and outsourced workforces are protected rather than used to conceal displacement.
Permit independent certification and remove members that repeatedly fail to meet the standard.
Invest in communities, not only individuals.
When AI-driven restructuring creates concentrated local job losses, fund regional economic diversification, small-business support, broadband, community colleges and public-interest technology projects.
Coordinate with local governments and worker organizations before closures or large reductions.
Avoid framing migration or individual reskilling as the only solution when an entire local labor market is affected.
Support a fair public-policy floor.
Advocate for unemployment insurance modernization, wage insurance, healthcare continuity, paid training, apprenticeship funding, worker-data protections and human review of consequential employment decisions.
Do not lobby for public subsidies for AI adoption while opposing the taxes, reporting duties or worker protections needed to manage its effects.
Treat collective action as a complement to enforceable law, not a substitute for it.
How to judge whether the commitment is real
A program is meaningful only if it is funded, measurable and enforceable. Useful indicators include:
| Measure | What responsible performance looks like |
| Notice | Consultation occurs before an irreversible deployment or staffing decision |
| Worker voice | Affected workers or representatives have documented influence and appeal rights |
| Redeployment | A high share of affected workers move to comparable or better roles |
| Earnings | Median pay and benefits are preserved during transformation |
| Transition support | Spending per affected worker and successful placement are publicly reported |
| Early-career access | Paid junior roles, apprenticeships and conversion rates do not collapse as AI adoption rises |
| Gain sharing | A defined share of AI-created value reaches workers or funds reduced hours and transition support |
| Job quality | Workload, autonomy, safety, surveillance and schedule stability improve or remain protected |
| Equity | Outcomes are tested for disproportionate harm across demographic groups and employment classes |
| Accountability | Independent reviewers can verify claims, and leaders face consequences for unmet commitments |
The ethical test is not whether a company can demonstrate that AI is efficient. It is whether the company can show that the people exposed to disruption had agency, received a fair share of the benefit, retained a realistic path to dignified work and were not abandoned when adaptation failed.
Q: How can society ensure that people can afford to live if AI reduces labor demand?
Answer: This is a distribution and governance question as much as a technology question. Productivity growth can support higher wages, lower prices, shorter working hours, better services or greater profits. None of those distributions occurs automatically. Historical automation research shows that job-replacing technology can widen wage inequality when displacement is not balanced by new tasks and opportunities.[15]
No single policy is sufficient, and several proposals below remain untested at the scale of a major AI transition. A resilient approach would combine adjustment support, worker power, education, competition and broad access to productivity gains.
1. Modernize income protection for displaced workers
Unemployment insurance, wage insurance and adjustment assistance can reduce income loss while people search, retrain or move into new work. Evidence from previous economic shocks shows that displacement can reduce earnings for years and that existing U.S. systems often reach too few workers or provide inadequate support. Policy research therefore recommends strengthening unemployment insurance while evaluating wage insurance and new forms of adjustment assistance.[16]
Support can include income replacement, healthcare continuity, paid retraining, job-search assistance and targeted help for communities experiencing concentrated losses. Program design must avoid forcing people into low-quality jobs merely to retain benefits.
2. Fund training tied to real work and demand
Effective adaptation requires more than telling individuals to reskill. Promising mechanisms include employer-supported training, paid learning time, apprenticeships, community-college pathways, portable credentials and wraparound support for childcare, transportation and connectivity. Training should be linked to verified labor demand and evaluated by completion, placement, earnings and job quality—not enrollment alone.[17][16]
3. Preserve entry-level career ladders
Because emerging evidence identifies reduced hiring of young workers as an early risk, employers, educational institutions and governments should preserve paid routes for acquiring experience. Apprenticeships, supervised junior roles, rotations and work-integrated learning can combine AI use with deliberate skill formation. The objective is not to prohibit AI from entry-level tasks; it is to ensure people still learn the tacit knowledge, judgment and accountability that senior work requires.[13][3]
4. Strengthen worker voice over AI deployment
Workers and their representatives should have meaningful input into how workplace AI changes staffing, workload, monitoring, data collection and performance evaluation. Policy options include notice and consultation, collective or sectoral bargaining, limits on intrusive surveillance, rights concerning worker data and accessible human review of consequential decisions. Research on inclusive AI policy identifies worker participation as important both for job quality and for finding productivity-enhancing uses that employees understand in practice.[16][17]
5. Share productivity gains more broadly
Possible mechanisms include stronger wage standards, profit sharing, employee ownership, tax-and-transfer reforms, broader capital ownership, guaranteed-income programs or public wealth funds. These are policy choices with different fiscal costs, incentives and political tradeoffs; the existing labor-market literature does not establish one as the proven response to future AI displacement. What is supported is the underlying risk: productivity gains will not necessarily reach workers without institutions that transmit them through wages, ownership, public services or income support.[16][18]
6. Consider shorter hours and work-sharing where productivity permits
If AI raises output per hour, firms and societies can choose to distribute part of the gain as reduced working time rather than only as reduced staffing or higher output targets. Four-day weeks, work-sharing subsidies and shorter standard hours are possible mechanisms, but evidence from shorter-workweek trials cannot yet establish their effectiveness as a response to large-scale AI displacement. Implementation must account for wages, coverage, scheduling, small-business costs and sectors that require continuous staffing.[18]
7. Regulate consequential AI decisions
AI used in hiring, worker management, lending, insurance, education, healthcare, housing, benefits, law enforcement, migration and critical infrastructure can affect rights and life opportunities even when it does not eliminate a job. Appropriate controls include risk and impact assessment, representative data testing, traceable logs, documentation, security and accuracy testing, competent human oversight, monitoring, incident reporting and accessible review or appeal. The exact legal requirements vary by jurisdiction; NIST provides voluntary U.S. risk-management guidance, while the EU AI Act creates binding requirements for defined high-risk systems on its implementation timetable.[11][14]
Q: How are governments responding to AI’s impact on work?
Answer: As of August 23, 2026, governments are beginning to respond to AI’s effects on work, but the response remains fragmented. Most U.S. federal initiatives described below are proposals rather than enacted protections, and their status may change.
Measurement and disclosure: The proposed AI-Related Job Impacts Clarity Act would require covered companies and federal agencies to report AI-related layoffs, hiring changes and retraining. The proposed AI Workforce Impact Study Act would require a GAO study of AI’s effects on employment, wages, occupations and demographic groups.
Worker-transition funding: The proposed National Workforce Transition Fund Act would support training accounts, tuition assistance, retention and redeployment grants and labor-market data improvements. The proposed AI Tax and Work Protection Act would tax certain large-scale AI activity and AI-associated workforce reductions to finance a federal worker-protection and job-creation program.
Workplace safeguards: Illinois law now prohibits discriminatory AI employment decisions and requires notice. New York City Local Law 144 requires bias audits and notices for covered hiring tools. California’s proposed SB 947 would require meaningful human review when automated systems contribute to discipline or termination. Other proposals would require advance notice and reskilling before technology-driven displacement.
European action: The EU AI Act classifies many employment AI systems as high-risk and will require risk controls, documentation and human oversight, although principal employment provisions are now scheduled for December 2027. The Platform Work Directive, which member states must transpose by December 2, 2026, adds transparency and human-review protections for algorithmically managed platform workers.
Enforceable U.S. protections currently concentrate on discrimination, notice and hiring-system audits rather than preventing job loss or guaranteeing economic security. No comprehensive federal framework yet guarantees income replacement, healthcare continuity, paid retraining, worker consultation or a share of AI productivity gains. Whether emerging proposals become meaningful protection will depend on enactment, funding, enforcement and whether they reach workers before displacement rather than only after it.
Q: What is a practical company commitment to protect people in the Age of AI?
Answer: A credible pledge could state:
When AI creates measurable value for our company, we will share that value with workers; assess and disclose foreseeable workforce impacts; consult affected people before significant deployment; invest in paid adaptation and entry-level opportunity; prefer redeployment and reduced hours over involuntary separation; provide meaningful income, benefit and career support when displacement cannot be avoided; and submit our outcomes to independent review.
Sources & validation notes
The research, institutional publications and official policy materials below support the FAQ answers. Publication details, links and legislative status were checked on August 23, 2026.
Validation note: Factual claims were cross-checked against the cited research, official institutional publications and legislative text. This is not the same as proving every conclusion: study findings remain limited by their methods, samples and dates. Anthropic’s analysis is company-authored, and the Stanford working paper documents an employment pattern without establishing cause and effect. The company, collective and public-policy recommendations—and the sample pledge—are normative proposals, not empirically validated outcomes.
Research and evidence
International Labour Organization — Generative AI and Jobs: A Refined Global Index of Occupational Exposure (ILO Working Paper 140, 2025).
OECD — Who Will Be the Workers Most Affected by AI? A Closer Look at the Impact of AI on Women, Low-Skilled Workers and Other Groups (OECD Artificial Intelligence Papers No. 26, 2024).
Anthropic — Labor Market Impacts of AI: A New Measure and Early Evidence (company-authored analysis, 2026).
International Labour Organization — Generative AI and Jobs: A 2025 Update (2025).
International Labour Organization — Generative AI and Jobs: A Global Analysis of Potential Effects on Job Quantity and Quality (ILO Working Paper 96, 2023).
Erik Brynjolfsson, Danielle Li and Lindsey R. Raymond — Generative AI at Work, The Quarterly Journal of Economics (2025).
Shakked Noy and Whitney Zhang — Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence, Science (2023).
Kevin Zheyuan Cui, Mert Demirer, Sonia Jaffe, Leon Musolff, Sida Peng and Tobias Salz — The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers, Management Science (2026).
Fabrizio Dell’Acqua and coauthors — Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality, Organization Science (2026; working-paper findings first released in 2023).
OECD — The Effects of Generative AI on Productivity, Innovation and Entrepreneurship (OECD Artificial Intelligence Papers No. 39, 2025).
National Institute of Standards and Technology — AI Risk Management Framework (AI RMF 1.0, 2023; subsequent updates noted on the source page).
Michael S. Barr, Federal Reserve Board — What Will Artificial Intelligence Mean for the Labor Market and the Economy? (2026).
Erik Brynjolfsson, Bharat Chandar and Ruyu Chen — Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence (Stanford Digital Economy Lab working paper; descriptive, not causal, 2026).
European Commission — AI Act: Regulatory Framework for Artificial Intelligence.
MIT News — Study Finds Stronger Links Between Automation and Inequality, reporting research by Daron Acemoglu and Pascual Restrepo (2020).
Urban Institute — Policy Priorities for an Inclusive AI Economy (2026).
International Labour Organization — Generative AI and Jobs: Policies to Manage the Transition (2023).
Xavier de Souza Briggs, Brookings Institution — Getting to All-of-the-Above: A Framework of Solutions for AI’s Coming Impacts on Work and Workers (2026).
Legislation and official policy materials
U.S. Congress — S. 3108, AI-Related Job Impacts Clarity Act (introduced).
U.S. Congress — H.R. 9427, AI Workforce Impact Study Act of 2026 (introduced).
U.S. Congress — S. 5055, National Workforce Transition Fund Act of 2026 (introduced).
U.S. Congress — H.R. 10044, AI Tax and Work Protection Act (introduced).
Illinois General Assembly — Public Act 103-0804, governing AI use in employment decisions.
New York City Department of Consumer and Worker Protection — Automated Employment Decision Tools and Local Law 144.
California Legislature — SB 947: Employment—Automated Decision Systems.
European Commission — EU AI Act.
EUR-Lex — Directive (EU) 2024/2831 on Improving Working Conditions in Platform Work.
Put Human Dignity at the Center of AI Progress.
Explore connected governance for AI-assisted work, autonomous Agents, Human decisions, assurance, evidence, and continuous improvement.