No Pay Rise, More AI: The Corporate Trade-Off Employees Are Being Asked to Fund
When Teradata employees were told they would not receive their customary annual salary increases in 2026, the company did not blame a recession, a collapse in sales or an unexpected financial emergency. Chief executive Steve McMillan told staff that money allocated to salary adjustments would instead be redirected towards artificial intelligence.
The message was unusually explicit. Teradata, a cloud analytics and data company employing approximately 5,100 people, wanted to increase its AI capabilities and recruit more specialist talent. Existing employees, except those in countries where increases were legally required, would help finance that ambition by accepting unchanged base salaries. Performance bonuses and equity awards would remain available, but the normal pay-review budget was being reallocated.
Companies make difficult investment choices every year. A decision to delay a factory, close an office or reduce a marketing budget may be commercially defensible. Choosing AI over employee pay is different because it turns a strategic technology investment into a direct statement about how management values labour. Employees are effectively being told that the company expects greater future returns from machines, models and a smaller group of AI specialists than from maintaining the purchasing power of its existing workforce.
That may prove correct in narrow accounting terms. It can still be damaging management.
A salary freeze is not a neutral decision
Executives often describe a pay freeze as maintaining compensation rather than cutting it. Employees experience it differently. When prices continue to rise, an unchanged salary buys less. Even modest inflation turns a nominal freeze into a real-terms reduction in pay.
There is also a cumulative effect. A missed increase does not disappear when the following year’s review arrives. Because future percentage rises are normally calculated from an employee’s existing salary, one frozen year can depress earnings for several years unless the company later makes a corrective adjustment.
The opportunity cost is therefore borne by the workforce immediately, while the benefits of AI investment remain uncertain and may accrue primarily to shareholders, senior management or newly hired specialists. Employees are asked to accept a guaranteed loss in purchasing power in exchange for a possible improvement in the company’s future competitiveness.
The psychological message may matter as much as the money. Pay is not only compensation for output; it is also one of the clearest signals an organisation sends about contribution, status and fairness. When a company publicly frames AI and employee raises as competing uses of the same money, it encourages workers to interpret the technology as their rival rather than their tool.
Management may subsequently promote AI adoption, experimentation and collaboration. Yet employees who have been told that AI has consumed their salary budget have a rational reason to distrust that invitation.
Why companies are making the trade-off
The pressure to spend is real. Companies are purchasing software licences, cloud capacity, specialist advice, data infrastructure and cybersecurity controls. They are also competing for engineers, product leaders and managers with credible AI implementation experience. For large technology groups, the investment extends to chips, networks, power supplies and data centres.
Businesses surveyed in late 2025 expected their average AI expenditure to rise from approximately 0.8 percent of revenue to 1.7 percent in 2026. At the largest technology companies, the sums are vastly greater. The global AI infrastructure cycle is now measured in hundreds of billions of dollars.
The chief financial officer therefore faces a familiar capital-allocation question: which existing expenditure can be reduced to finance the new priority? Labour is one of the largest costs in many businesses, making compensation, hiring and headcount obvious targets.
This can create a compelling spreadsheet. A company freezes salaries, leaves vacant positions unfilled and uses the savings to purchase AI tools. Employees equipped with those tools supposedly produce more work, enabling the organisation to grow without restoring the removed labour costs.
The difficulty is that the productivity assumption is frequently treated as established before the company has demonstrated it in its own operations.
Recent research by the Federal Reserve Bank of Atlanta found that corporate executives attributed an average labour-productivity gain of 1.8 percent to AI investment in 2025 and expected this to reach 3 percent in 2026. These are meaningful gains, but they are considerably more restrained than the dramatic figures routinely quoted in consultancy marketing.
An analysis of more than 12,000 European and American companies by the European Investment Bank found an approximately 4 percent increase in labour productivity among AI adopters. Importantly, the result appeared to come from capital deepening rather than widespread job destruction. The productive company was not simply substituting software for people; it was giving workers better capital with which to perform their jobs.
That distinction matters. AI may improve employee output, but that does not prove that suppressing employee pay is the best way to finance it.
The most productive AI companies are not necessarily paying less
The emerging labour evidence complicates the idea that companies must choose between artificial intelligence and employees.
PwC’s 2026 analysis of more than one billion job advertisements found that productivity growth was substantially stronger in industries most exposed to AI. Yet those industries were also recording faster wage and employment growth than less-exposed sectors. The apparent leaders were not treating labour as a residual cost to be squeezed after the technology budget had been set. They were combining new tools with higher-value skills.
This is economically plausible. AI rarely delivers value merely because a company has purchased access to a model. Employees must identify suitable tasks, prepare and protect data, redesign workflows, verify outputs and recognise when automated recommendations are wrong. Experienced staff also possess the operational knowledge needed to distinguish a technically impressive demonstration from a useful business process.
An insurer cannot safely automate claims work without people who understand policy wording, fraud patterns, regulatory obligations and unusual cases. A bank cannot delegate credit decisions simply because a model can summarise financial documents. A manufacturer still needs employees who know where delays, quality failures and safety risks actually arise.
Removing incentives from these people while investing in the tools that depend on their knowledge is a questionable form of transformation. The company may save on compensation while weakening the institutional capacity required to make the AI useful.
The risk is particularly acute when an employer freezes the pay of its established workforce but offers premium packages to recruit scarce AI specialists. Incumbent employees see a company that claims to lack money for their progression while finding it for external hires associated with the fashionable strategic priority. The resulting resentment is not resistance to innovation. It is a predictable reaction to unequal treatment.
AI investment can become a cover for ordinary cost-cutting
Not every project described as an AI investment represents genuine transformation. The label can provide a more attractive explanation for cost reduction than weaker demand, poor financial planning or pressure from investors.
A company announcing that it is “becoming AI-first” appears forward-looking. A company admitting that it is reducing labour costs to protect margins appears defensive. The distinction matters because the first narrative may be used to justify actions that management wanted to take regardless of the technology.
This does not mean that AI-related restructuring is fictitious. Oracle, for example, reduced its workforce by about 21,000 during its 2026 financial year while simultaneously pursuing an exceptionally capital-intensive expansion of its AI cloud infrastructure. The company is making a material shift in where it allocates capital.
But managers should be able to show a credible connection between the expense being removed and the value the new system is expected to create. “We need to invest in AI” is not an investment case. It is a strategic preference.
A serious proposal should identify the business process being changed, its current cost and performance, the expected improvement, the implementation expense, the operational risks and the period over which returns will be measured. It should also explain whether the benefit depends on employees doing more work, fewer employees performing the same work, or the company creating a product or service it could not previously offer.
Without that level of specificity, employees may reasonably conclude that AI is being used as a convenient justification for a general pay restraint.
What a credible AI investment case should contain
Before funding AI through compensation reductions, a board should require management to answer several practical questions.
The first is whether the company has proved that the proposed use case works. A controlled pilot should compare the AI-assisted process with the existing one using measures such as completion time, error rates, customer outcomes, revenue or cost per transaction. Time saved is not automatically value created. If employees use the saved time correcting unreliable outputs or attending additional meetings, the apparent efficiency may never reach the income statement.
The second is whether the full cost has been calculated. Licence fees are only one component. Integration, data preparation, security, legal review, employee training, monitoring, model usage and human supervision can make deployment considerably more expensive than the initial proposal suggests.
The third is who will capture the productivity gain. An employee asked to process 20 percent more work with AI but denied a salary increase experiences the programme as work intensification, not augmentation. Companies do not need to distribute every efficiency gain immediately, but some credible sharing mechanism can improve adoption. This might include bonuses tied to measurable gains, additional development opportunities, reduced administrative workloads or a commitment to reconsider salaries once agreed milestones are achieved.
The fourth is what happens when the tool fails. AI systems can produce incorrect answers, mishandle sensitive information and behave unpredictably when connected to other software. A company reducing experienced staff at the same time as it increases reliance on immature systems may remove precisely the people capable of recognising those failures.
Finally, management should state when it will judge the investment unsuccessful. Projects without defined stop criteria can continue absorbing money because ending them would require executives to admit that an important strategy has not worked.
How the decision should be communicated
There is no communication technique capable of making a pay freeze popular. Management can, however, avoid making the situation worse.
The first requirement is precision. Leaders should say how much is being invested, what the money will purchase, which business outcomes are expected and when the decision will be reviewed. Vague language about transformation, agility and winning with AI will sound evasive when employees are being asked to make a tangible sacrifice.
Executives should also explain what they themselves are giving up. Freezing ordinary salaries while preserving senior bonuses, large equity awards or discretionary executive benefits creates an obvious fairness problem. Even when those payments are governed by existing contracts, leadership must recognise how the contrast will be perceived.
The company should not promise that AI will never affect jobs unless it can substantiate that claim. Nor should it insist that employees enthusiastically embrace the technology. A more credible message acknowledges uncertainty, specifies which tasks are likely to change and offers paid time for relevant training.
Above all, employers should avoid claiming that the decision is good for employees simply because a more competitive company may offer better long-term security. That may be management’s genuine belief, but it does not remove the immediate transfer of risk. Workers are still surrendering compensation today to fund an investment whose eventual returns they do not control.
The false choice between technology and people
Companies must invest in AI where it can improve a real product, process or decision. Refusing to do so because employees dislike change would be irresponsible. But presenting technology and fair compensation as natural competitors is equally shortsighted.
The firms most likely to generate lasting value from AI will need more than software. They will need people who understand customers, operations, risk and the limits of automated systems. They will need employees willing to expose inefficient processes, test unfamiliar tools and take responsibility when an agent’s recommendation cannot simply be accepted.
A salary freeze may release cash quickly. It can also produce disengagement, departures and quiet resistance that never appears in the original investment model. Replacing those employees later, particularly once their knowledge has left with them, may cost more than the pay increases the company initially withheld.
The real test is therefore not whether management can finance an AI budget by reducing compensation. It is whether the resulting organisation is more capable of turning that technology into value. When employees are treated as the source of funding rather than the means through which transformation will occur, the answer may be no.
