Consider a team that introduces AI and begins delivering work faster. Output rises, employees take on more complex tasks and some assignments require fewer hours. Then comes the appraisal cycle: who deserves a higher rating, a larger bonus or a different pay position?
The employee producing the most may not hold the scarcest skill. The colleague checking the work may prevent costly errors without appearing more productive. Someone else may have improved the workflow for the entire team, while employees with less access to tools or training face changing expectations.
For CHROs and Total Rewards leaders, the dilemma is not whether AI-assisted performance deserves a reward. It is whether existing systems can explain whose contribution they are rewarding, and why.
Higher output complicates the contribution question
AI makes familiar productivity measures harder to interpret. More reports, faster resolutions or additional code can reflect useful progress, but volume alone reveals little about quality, rework, customer value or risk.
Research published in The Quarterly Journal of Economics in 2025 illustrates the variation. Among 5,172 customer-support agents, access to a generative AI assistant increased issues resolved per hour by 15% on average. Less-experienced and lower-skilled workers improved both speed and quality, while the most experienced workers saw smaller speed gains and some quality declines. The findings concern a specific setting, not a universal productivity benchmark.
For employers, this raises a measurement problem before it becomes a compensation problem. Does an existing target still capture valuable performance once the tools change?
Effective AI use is itself a contribution. The challenge is recognising it without treating every increase in output as equivalent, or preserving manual effort as proof of merit.
A scarce skill and strong performance are not the same thing
PwC’s 2026 Global AI Jobs Barometer, based on more than one billion job advertisements across 27 countries and territories, reports an average 62% wage premium for AI skills, compared with 57% in the previous edition. This compares advertised salaries for roles requiring AI skills with comparable roles that do not; it is neither an automatic individual pay rise nor an India-specific benchmark.
The distinction matters inside organisations. Market scarcity can explain why a capability commands higher pay, but it does not establish how well an employee applies it. Equally, someone in a less scarce role can make an exceptional contribution without commanding the same market premium.
A rewards system must accommodate both realities. Otherwise, organisations risk confusing what they must pay to secure talent with what they should reward for performance.
In his pre-TRWC conversation with People Matters, RPG Group’s Debashish Majumdar places the emphasis on outcomes employees can influence, the intelligent use of AI and the judgement they bring to the work.
What should become permanent pay?
An emerging skill creates an immediate attraction problem, but the compensation decision can have consequences long after the initial shortage passes.
Where AI changes a role’s enduring responsibilities, complexity or accountability, job evaluation and base-pay positioning may need to change. Where the need is tied to a defined assignment or temporary scarcity, a separately identified, time-bound premium may be more appropriate.
Neither approach should become a default simply because the skill carries an AI label. Leaders need clarity about what the payment recognises, how application is demonstrated and when the arrangement is reviewed.
They also need to consider the combined proposition. A hiring premium, skills allowance and performance bonus may each have a rationale, yet collectively reward the same capability several times.
Benefits face their own relevance test
The implications extend beyond salary. As work changes, employees may need different forms of support to remain effective and see a future within the organisation.
Someone moving into a redesigned role may value funded learning, protected practice time and access to internal opportunities. A caregiver may prioritise dependent-health coverage or predictable flexibility. Another employee may need stronger long-term financial support.
These needs do not map neatly onto job titles, generations or performance categories. Employability support should not become a privilege reserved for the specialists already commanding a premium.
Choice can improve relevance, but it does not automatically reduce costs. Administration, employee understanding, service quality and actual uptake all affect value. Benefits should be assessed against the needs they address, rather than the length of the options menu, and should not substitute for fair pay or credible progression.
Junior talent changes the equation
The early-career question is not simply whether AI removes routine tasks. It is whether employees are being expected to exercise judgement before they have had sufficient opportunity to develop it.
An October 2026 working paper published through Stanford’s Digital Economy Lab studied 673 final-year IT apprentices in Germany. AI access improved performance on tasks beyond their formal training without reducing immediate task comprehension. The study does not establish long-term expertise, but it challenges the assumption that AI-assisted output necessarily means learning has been bypassed.
The organisational test is whether assisted performance becomes transferable capability. Can employees explain a recommendation, identify a flawed response and handle an unfamiliar exception?
Progression needs to capture that development. Experienced colleagues also need recognition for mentoring and knowledge-sharing; otherwise, individual output targets can make building future capability someone else’s unrewarded responsibility.
Differentiation needs an explanation people can question
Employees do not need identical rewards to be treated fairly. They need a defensible explanation of the differences and a credible opportunity to challenge them.
That starts before appraisal. Did employees have comparable access to approved tools, relevant training and opportunities to demonstrate new capabilities? Did the assessment recognise shared expertise, or attribute a team improvement entirely to the person most visible on the dashboard?
The use of time saved also matters. If every efficiency gain immediately becomes a higher target, with no room for learning or recovery, productivity can become an intensification of work rather than a better employee proposition.
Before the next rewards cycle, leaders should be able to explain what their incentives now measure, why particular premiums exist and how employees can progress. If those explanations remain unclear, sharper differentiation may create more questions than confidence.
Bring the dilemma to TRWC 2026
AI is evolving faster than work. It is testing the relationship between market value, individual contribution, employee support and organisational fairness—the connections running through TRWC’s Impact, Choice, Trust and Energy pillars.
Join the HR leaders at People Matters TRWC 2026 on 18 November at The Leela Ambience, Gurugram, under the theme “Orchestrating Performance in a Choice-Driven Workforce.”
