Oliver Wyman AI Job Security Survey Finds Indian Workers Would Trade Pay for Protection

Oliver Wyman’s AI job security survey found that 79% of employed Indian respondents would trade pay for protection from AI replacement.

The respondents said they would surrender at least 10% of their salaries for that guarantee. Only 47% of employed respondents globally expressed the same willingness, according to the report.

That gap makes the India result more than another measure of abstract AI anxiety. Workers are assigning a concrete financial value to employment security while adopting AI rapidly in their jobs and personal lives.

The central conflict is between the productivity employers expect and the employment risk workers believe they will absorb. Oliver Wyman describes AI as creating capacity, but employees expect companies to turn some of that capacity into smaller teams.

This expectation is not yet a measurement of actual displacement. It is evidence of a trust problem that can shape adoption, retention, governance, and workforce planning before layoffs appear in official data.

The Oliver Wyman AI Job Security Survey Puts a Price on Fear

Indian respondents did not merely say that AI worried them; most accepted a hypothetical pay cut for protection from replacement.

The result comes from India-specific analysis of the Oliver Wyman Forum Global Voices Survey. The organization released its broader AI reality check on October 15, 2026.

The research included more than 25,000 responses collected across 15 markets. Two survey waves were conducted in April and August 2026.

Oliver Wyman also used responses from 415 chief executives in its 2026 CEO Agenda research. Those executive responses informed the wider analysis of investment, organizational change, and AI returns.

The headline result concerns employed respondents in India. Seventy-nine percent would surrender at least 10% of their salaries for a guarantee that AI would never replace them.

The comparable global share was 47%. Across all markets, 68% would give up some portion of their salaries to join an employer promising never to replace people with AI.

These questions describe hypothetical choices rather than negotiated employment contracts. Nobody in the survey necessarily received an actual offer pairing lower pay with an enforceable job guarantee.

Still, the threshold makes the result unusually tangible. A respondent accepting a 10% reduction is describing a substantial perceived cost, not expressing mild concern about technological change.

The report also asked what employees thought would happen if AI doubled their teams’ output. Workers ranked smaller teams last among their preferred outcomes but first among their expected outcomes.

That reversal reveals the deeper issue. Employees hear productivity language, yet many interpret it as a warning that employers will eventually need fewer people.

Oliver Wyman summarizes this tension through the statement that workers will not embrace a future they believe excludes them. The survey suggests that this belief already affects how employees value compensation and security.

The India figure must remain in context. The public report identifies the global response total and markets, but its public summary does not provide India’s exact subsample size.

It also does not establish how many employers could legally or financially provide permanent protection from AI-related replacement. The finding measures preferences under a hypothetical guarantee, not the feasibility of that guarantee.

Even with those limits, the comparison matters. Indian respondents expressed considerably stronger demand for employment protection than the wider international sample.

That response appeared alongside high enthusiasm for AI in other parts of life. It therefore cannot be reduced to general opposition toward the technology.

India’s AI Job Displacement Anxiety Coexists With Fast Adoption

The striking part of India’s result is that fear rises beside AI adoption, not in place of it.

The same research found that 59% of Indian respondents would consult AI first when searching for a new product or service. The global comparison was 33%.

Six in ten Indian respondents would follow an AI recommendation over advice from friends, relatives, social media, or online reviews when those sources conflicted. The worldwide share was 32%.

Another 81% would consider switching brands if an AI agent found a better match. Globally, 60% expressed that willingness.

Seventy-two percent of respondents in India would let an AI agent spend money for them without checking first. The corresponding global result was 49%.

These findings depict unusually high confidence in AI as a consumer decision tool. They also make the employment result harder to dismiss as technological unfamiliarity.

People can trust an AI system to select a product while distrusting an employer’s plans for the same technology. Consumer adoption and workplace confidence are separate judgments.

A shopper can reject a bad recommendation, change services, or decline a purchase. An employee usually has less control over job redesign, staffing decisions, performance measurements, or automation targets.

The survey also found that 79% of daily AI users in India had downplayed their AI abilities at work. The global result was 62%.

That behavior seems contradictory at first. Workers often hear that AI proficiency will improve their careers, yet some avoid advertising the very skills employers request.

One possible explanation is defensive behavior. An employee may worry that revealing a faster AI-assisted workflow will lead to more work, higher targets, or evidence that fewer colleagues are needed.

Oliver Wyman’s global findings support that concern without proving the explanation for every respondent. Only 8% of workplace AI users reported a company-wide standard for allocating AI-saved time.

Another 54% said individuals decide what happens to that time, or no explicit decision exists. Productivity becomes visible, but its rewards and ownership remain unclear.

That uncertainty can encourage workers to keep improvements private. A person who saves several hours may quietly use the time rather than invite management to redefine the job.

This dynamic can also weaken measurement. Executives may know that employees use AI but remain unable to connect scattered time savings with revenue, service quality, or operational costs.

John Romeo, chief executive of the Oliver Wyman Forum, described the problem through company financial results. AI creates capacity, he argued, but leaders must decide how to use that capacity.

For workers, the unresolved question is more personal. Does the new capacity create room for better work, or does it become justification for fewer positions?

The willingness to exchange salary for security represents one answer. Many respondents value a clear employment commitment more than the uncertain upside attached to AI-driven productivity.

The Productivity Promise Collides With the Smaller-Team Expectation

Employers present AI as augmentation, while many workers expect its gains to support headcount reduction.

Augmentation means using technology to expand a person’s capabilities rather than eliminate the role. Substitution means transferring tasks, responsibilities, or entire positions from people to automated systems.

Most workplace deployments contain both. An AI assistant can reduce drafting time while leaving judgment, accountability, and client relationships with the employee.

The balance becomes politically important when companies discuss productivity. A percentage improvement sounds attractive to investors, but workers immediately ask who receives the benefit.

Employees might receive shorter hours, higher compensation, better assignments, or more time for customers. Employers might instead raise output expectations, slow hiring, or eliminate roles.

Oliver Wyman found a sharp difference between preferred and expected outcomes. Employees placed workforce reduction last when choosing how doubled output should be used.

When predicting their employers’ response, they placed smaller teams first. That gap between desired and expected outcomes is the article’s primary conflict.

It also places managers under pressure. Leaders cannot ask employees to experiment openly while leaving the employment consequences deliberately vague.

Clear communication alone will not resolve the issue. Workers will compare corporate language with hiring decisions, restructuring plans, workload changes, and promotion opportunities.

The survey found that only 37% of white-collar executives believed leadership had set a clear AI ambition and communicated progress. Just 11% of non-managers agreed.

That divide suggests employees and senior leaders experience the same program differently. Executives see portfolios, budgets, pilots, and strategic goals.

Individual contributors see new tools entering specific workflows. They also see vacancies left unfilled, entry-level tasks disappearing, or performance expectations changing.

Middle managers occupy the difficult space between those perspectives. They must implement new systems while answering questions that senior plans may not address.

The training pathway is also changing. Seventy-four percent of white-collar respondents at middle-management level or above said AI performed tasks once used to train junior employees.

In India, 45% of white-collar managers said AI performed much of the work from which junior employees previously learned. The global comparison was 25%.

This shift creates a quieter form of displacement. A company might retain its existing workforce while weakening the route through which new workers acquire experience.

Entry-level work often includes drafting, research, documentation, analysis, and routine customer support. Those tasks can provide the context needed for later judgment-heavy responsibilities.

If AI absorbs them, companies need a replacement training model. Otherwise, they risk preserving current expertise while reducing the pipeline that produces future specialists and managers.

The broader report found that 80% of daily AI users learned at least as much from AI as from their managers. Forty-seven percent said AI taught them more.

That finding shows why workers continue adopting the technology despite their fears. AI can function as both a useful instructor and a potential substitute.

The tension is therefore not humans against machines. It is employer promises of augmentation against worker expectations of workforce reduction.

AI Workplace Anxiety Is Real, but Displacement Evidence Remains Mixed

A willingness to sacrifice salary measures perceived risk, not the number of jobs that AI has eliminated.

Survey responses can move before labor-market outcomes. They can also reflect news coverage, employer messaging, personal AI use, and wider economic uncertainty.

Recent job-loss research from Stanford researchers provides an important counterweight. It examined worker beliefs, AI adoption, job postings, and layoffs in the United States.

The researchers estimated that 30% to 40% of American workers used generative AI through the first half of 2026. The average worker perceived a 20% chance of losing a job to it within two years.

However, the study found no statistically significant response in job postings or layoffs across more AI-exposed occupations. Fear had moved ahead of observable displacement in those measures.

The researchers found that concern increased with usage intensity. It was also stronger among people applying AI to core occupational functions, including programming.

That pattern resembles the contradiction in the Indian findings. Greater familiarity can expose the technology’s usefulness while making its substitution potential easier to imagine.

Research from the Boston Fed also distinguishes personal concern from expected industry disruption.

Its worker survey found that just over 10% feared losing their own jobs to AI at the end of 2025. That share had almost doubled from 5% one year earlier.

Yet 60% expected AI-related layoffs or a reduction in employment across their industries. Workers often see disruption as likely while considering their own positions relatively safe.

The same research found that AI can replace individual tasks without removing whole occupations. Employees may become more productive because technology handles selected portions of their work.

That task-level view helps explain why current employment data and worker anxiety can diverge. An occupation can remain intact while its staffing needs, career ladder, and daily responsibilities change.

The Oliver Wyman finding also relies on a hypothetical guarantee with unusually strong wording. Permanent protection from replacement removes uncertainty that no ordinary employment arrangement can eliminate.

Respondents may therefore be valuing predictability more broadly. Their answers could incorporate fear of layoffs, deskilling, stagnant promotion, rising targets, or unwanted role changes.

The exact wording still matters because the offer requires a minimum 10% salary sacrifice. The result indicates that uncertainty carries a high perceived cost for many Indian respondents.

It does not show that 79% expect immediate job loss. It also does not show that they would accept a real contract after reviewing its conditions.

Cross-country comparisons require similar caution. Salaries, labor protections, household savings, informal employment, and career mobility differ considerably between markets.

A 10% pay reduction does not create the same household impact everywhere. Neither does a promise of job security have equal credibility under every legal system.

The public materials also do not provide margins of error for the India-specific percentage. Readers should treat the number as a strong sentiment signal rather than a precise forecast.

Independent research nevertheless supports the direction of concern. Workers across several markets increasingly expect AI to reduce employment or change the structure of their industries.

The unresolved question is speed. Available evidence shows rapid task adoption, growing fear, and changing workflows, but less proof of immediate economy-wide job elimination.

Governance Problems Turn Adoption Into a Trust Test

The survey connects job insecurity with a second problem: employees are adopting AI faster than organizations are governing it.

Seventy-three percent of workplace AI users in India reported entering sensitive company information into an AI tool. The global comparison was 50%.

Sensitive information can include internal documents, customer details, strategy material, or proprietary data. The public summary does not specify the exact information submitted by each respondent.

Another 62% of Indian respondents said inaccurate AI-generated information caused a real problem during the previous year. Globally, 46% reported such an experience.

These figures show that employees are not simply observing AI from a distance. Many use it in consequential workflows despite uncertain rules and known reliability problems.

The broad report found higher reported productivity when approved AI was accessible, governed clearly, and integrated into daily software. Under those conditions, 88% of users reported productivity gains.

Where those conditions were mostly absent, 52% reported gains. Governance appeared to support adoption rather than merely constrain it.

Rules can still become counterproductive. Sixty-three percent of employees who considered their organizations’ policies too strict used unapproved AI tools.

The equivalent share was 38% when employees considered the approach appropriate. Excessive restrictions can push activity outside approved systems without reducing demand.

That creates a difficult management problem. Organizations need safeguards, yet those safeguards must accommodate the reasons workers use AI in the first place.

Job insecurity complicates compliance. Employees may avoid disclosing effective workflows if they believe transparency threatens their positions or increases their workloads.

A company can then lose visibility into both productivity gains and data risks. Its official AI program tells one story while informal use creates another.

Trust therefore becomes operational infrastructure. Workers need credible answers about saved time, performance measurement, data handling, and the future of their roles.

A generic statement that AI will assist employees is insufficient. Teams need rules covering which tasks can be automated and which decisions remain under human control.

They also need escalation paths for inaccurate output. Human review must remain meaningful when errors affect customers, regulated decisions, or proprietary information.

Companies should define how productivity gains will be shared. Options include improved service, reduced backlogs, new products, training time, and more manageable workloads.

These choices do not guarantee employment. They do let employees judge whether an AI program has purposes beyond cost reduction.

Workers also retain some agency. Building a personal knowledge system can preserve context, decisions, and expertise as tools and roles change.

That practice should not shift responsibility away from employers. Organizations control deployment, performance systems, staffing, and access to workplace data.

The India findings show what happens when adoption outruns those institutional choices. Employees use AI extensively while pricing protection against the consequences.

What Employers and Workers Should Watch Next

The next phase will be judged through staffing decisions, career development, and enforceable workforce commitments rather than adoption statistics alone.

The first signal is whether large employers publish clear policies for AI-created capacity. Those policies should explain what happens when teams complete the same work with fewer hours.

A policy that directs savings toward growth, service quality, or reduced workloads would weaken expectations of automatic workforce reduction. Silence would reinforce them.

The second signal is entry-level hiring and training. Oliver Wyman’s results indicate that AI already performs tasks traditionally used to develop junior employees.

Employers should disclose how apprenticeships, reviews, supervised work, and progression will change. A shrinking training pipeline would make the displacement concern more concrete.

A redesigned pipeline could point toward augmentation instead. Junior employees might supervise AI output, work with customers earlier, and receive structured practice in judgment-heavy tasks.

The third signal is the difference between announced layoffs and measured labor-market changes. Company statements often cite several reasons for restructuring, making AI’s contribution difficult to isolate.

Researchers will need occupation-level hiring, vacancy, wage, and separation data. Task exposure matters because job titles can remain stable while the content of work changes.

The Stanford findings currently weaken claims of widespread immediate displacement in the United States. Future evidence of falling demand in highly exposed occupations would strengthen the worker concerns captured in India.

Conversely, stable hiring combined with changing task profiles would support a restructuring interpretation. AI would alter work significantly without eliminating employment at the predicted scale.

Workers should also watch how companies reward AI proficiency. If greater skill produces higher targets without better compensation or advancement, employees will keep hiding useful practices.

If AI proficiency leads to stronger mobility and better assignments, openness becomes more rational. The incentive design will shape whether adoption stays informal or becomes institutional knowledge.

Security promises deserve scrutiny as well. A broad pledge never to replace people with AI sounds reassuring but can hide several unresolved questions.

Does the promise cover contractors, vacancies, and outsourced roles? Does it prevent layoffs, or only prohibit naming AI as the reason?

Does it protect a particular job, employment somewhere within the company, or only a period of retraining? The details determine whether workers receive genuine protection.

Governments and labor organizations may also enter the debate. They can require consultation, transparency, training, or impact assessments without attempting to freeze every job description.

Such interventions face their own tradeoffs. Overly rigid rules can slow useful adoption, while weak rules leave employees carrying uncertainty they did not create.

India has a particularly important role because its service economy contains many technology-enabled and globally delivered jobs. Those roles can benefit from AI while facing task-level automation.

The Oliver Wyman AI job security result captures that dual exposure. Respondents appear ready to use AI broadly, yet many want assurance that adoption will not make them expendable.

Employers now need to convert productivity claims into credible workforce plans. That means publishing what happens to saved time, protecting training pathways, and measuring outcomes beyond headcount.

Workers should ask equally concrete questions. Which tasks are changing, who reviews automated output, how will performance targets move, and what happens to people whose roles shrink?

The 79% figure should not be read as a prediction that four in five Indian workers will lose their jobs. It should be read as a warning about perceived bargaining power.

Employees believe security has become valuable enough to purchase with a meaningful share of their income. That belief can influence retention and workplace behavior before any displacement appears.

The decisive test will not be whether another survey produces a higher percentage. It will be whether employers make AI-created capacity beneficial and visible to the people creating it.

If staffing falls while output rises, the fear behind the survey will look prescient. If companies preserve careers and rebuild training, the willingness to trade pay may recede.

For now, every workplace AI proposal should answer one practical question: what does the employee receive when the technology makes the team more productive?