AI’s Long-Term Effects for Retirement Plans, Fiduciaries Unknown

AI is changing the practice of tax law. This series examines the ethical, legal, and practical implications of AI across key areas of tax practice.

This is a two-part article on the current and possible future effects of the AI boom on single-employer, multiple-employer, and multiemployer retirement plans.

Artificial intelligence is reshaping retirement plans. This article, the second installment of a two-part article, focuses on the longer-term effects of AI on retirement plans and how AI-driven changes in hiring, job stability, wages, and retirement timing may affect contribution patterns, benefit accruals, and plan funding across single-employer and multiple-employer arrangements. Part 1 of the two-part installment focused on AI’s already visible effects on retirement plans, particularly in plan administration, fiduciary processes, and cybersecurity.

The more durable effects of the AI boom are likely to arrive through labor markets before they arrive through plan design. Carlo Pizzinelli and Marina Tavares, writing for the Pension Research Council at Wharton, conclude that older workers generally have lower labor-market fluidity. They are less mobile across jobs, occupations, and industries and less resilient after job loss, which may limit their ability to adjust to AI-driven structural change.

But the same paper finds that a larger share of older workers is already employed in occupations expected to benefit from AI as a complementary technology and suggests that, conditional on acquiring the necessary skills, these workers may enjoy growing opportunities, higher wages, and stronger incentives to remain employed longer. Carlo Pizzinelli and Marina M. Tavares, AI and the Future of Work in an Aging Economy, Pension Research Council Working Paper No. WP2025-14, at 1-3, 11, 16, 30-31 (July 7, 2025). The central implication is that AI is unlikely to produce one retirement trajectory. It is more likely to widen the gap between workers whose late-career productivity rises with AI and workers whose bargaining position weakens because AI substitutes for what they do.

That matters first for single-employer defined contribution plans. If AI allows some employees to remain productive later in life, work more flexibly, and delay retirement, those employees may contribute for longer periods, postpone distributions, and accumulate larger balances.

If AI instead destabilizes other jobs, those workers may experience more leakage, interrupted savings, or involuntary earlier retirement. Rosa Aisa and Josefina Cabeza conclude in Research in Economics that workers who use AI in their jobs and derive the greatest benefit from it are more likely to extend their working lives, while workers who don’t use AI retain a more traditional retirement pattern. Rosa Aisa and Josefina Cabeza, Artificial Intelligence: Redefining the Retirement Pattern, 79 Research in Economics, no. 3, art. 101062 (2025).

Their model therefore points to divergence, not a single retirement effect.

A separate line of retirement research reinforces the same point from a different angle. In Work, Aging and Retirement, Elissa El Khawli, Mark Visser, and Mustafa Firat find that older workers with trajectories marked by stable low recognition and stable low resources face a higher risk of earlier retirement than workers with stronger and more stable job resources. Elissa El Khawli, Mark Visser and Mustafa Firat, Trajectories of Job Resources and the Timing of Retirement, 11 Work, Aging and Retirement 149 (2025).

That isn’t an AI paper, but it matters here because AI won’t land in a social vacuum. It will operate inside workplaces with different levels of autonomy, skill development, recognition, and security. Where AI supports resource-rich jobs, it may lengthen working lives. Where it intensifies surveillance, compression, or loss of recognition, it may push retirement earlier or make late-career employment less sustainable.

For single-employer defined benefit plans, the effect may appear less in interfaces than in assumptions. The scholarship doesn’t yet say in so many words that AI will force actuarial assumption changes. But it does suggest that possibility.

If AI changes retirement timing, late-career wages, turnover, job quality, or the distribution of longer versus shorter careers inside a workforce, those shifts could eventually reach the liability side of a pension plan. A reasonable inference is that corporate pension sponsors should treat AI not only as an administrative technology question but also as a workforce-composition question with potential consequences for retirement-age assumptions, benefit commencement patterns, and the future shape of liabilities.

There is also an asset-side channel. Pizzinelli and Tavares observe that if AI boosts firm productivity, expectations of higher future profits can raise stock values and dividend payouts, and they note that this channel may be particularly relevant for older workers, who on average own larger amounts of equity assets. Carlo Pizzinelli and Marina M. Tavares, AI and the Future of Work in an Aging Economy, Pension Research Council Working Paper No. WP2025-14, at 26-27 (July 7, 2025).

For retirement plans, that suggests the AI boom may first appear through familiar public-market exposures rather than through specialized AI products. Single-employer 401(k) participants already own significant public-equity exposure through broad equity funds and target-date structures.

Defined benefit plans likewise see changes in equity-market valuations through public-equity and growth-sensitive allocations. A reasonable inference is that one early investment effect of AI on retirement plans could be greater concentration and valuation sensitivity inside ordinary vehicles, rather than novelty risk arising only from specialized AI products.

Multiple-employer arrangements, especially pooled employer plans, or PEPs, may eventually become a practical channel for scaled AI-assisted retirement guidance. Chet Bennetts and Eric Ludwig report in the Financial Planning Research Journal, drawing on a survey of 2,000 public-sector employees, that workers with the highest AI adoption rates also showed the highest engagement with financial professionals—72% percent engagement among high adopters versus only 15% percent among the most technology-resistant segment—undercutting the assumption that AI simply displaces human advice.

Their study identifies five distinct adopter segments and further highlights an “Employer-Driven AI Users” segment, representing roughly 28% percent of respondents. This suggests that workplace-sponsored tools can draw more reluctant users into AI-assisted guidance, although the survey population was public-sector rather than private-sector workers.

That complements Andrew Lo and Jillian Ross’ broader scholarly argument in Harvard Data Science Review that practical financial-advice systems built on large language models still face three overarching challenges: domain-specific expertise tailored to the user’s situation, trustworthiness and adherence to ethical standards, and conformity to regulatory oversight. Chet R. Bennetts and Eric Ludwig, The Desire for AI Advice in Retirement Plans: A Latent Class Analysis, 11 Fin. Planning Research J., no. 2 (2025); Andrew W. Lo and Jillian Ross, Can ChatGPT Plan Your Retirement?: Generative AI and Financial Advice, Harvard Data Science Review (Aug. 6, 2024).

This doesn’t indicate that PEPs will replace advisers with chatbots. It means centralized plans may be well positioned to deploy AI as a supervised layer around human advice, segmentation, and participant support.
For multiemployer plans, the long-term AI effect may be the most indirect and the most serious. Morgan Lewis argues that AI will materially change the speed, scale, and analytical depth of labor strategy, especially in workforce intelligence, bargaining strategy, compensation monitoring, and employee-engagement tools.
Lowenstein and Proskauer, by contrast, remind the reader that multiemployer plans are funded through employer contributions tied to collectively bargained work and that reductions in contributions can generate substantial liability. Harry Johnson, III, Nicole Buffalano, Kelcey Phillips, and John Ring, How AI Will Fundamentally Reshape Work in Labor Relations, Morgan Lewis (Mar. 20, 2026); Andrew E. Graw, Taryn E. Cannataro, and Jessica I. Stewart, Multiemployer Pension Plans: Mitigating Risk in the Context of a Business Transaction, Lowenstein Sandler (Feb. 22, 2024;Proskauer Rose, Withdrawal Liability—What It Is and Why It Matters, Proskauer Benefits Brief: Legal Insight on Compensation and Benefits (Jan. 13, 2026).

That means the major future AI question for multiemployer plans is not whether trustees adopt sophisticated participant tools. It is whether AI changes the volume, location, classification, and bargaining value of covered work. If it does, it’s reasonable to infer that those changes could flow into contribution patterns, funding pressure, and withdrawal exposure.

The broad lesson across all three plan structures is that the AI boom is likely to divide into two phases. The first phase is already here and is dominated by fiduciary process, vendor oversight, cybersecurity, and recordkeeping discipline. The second phase is coming and will be driven by how AI changes work itself: who stays employed longer, who exits earlier, which employers can spread costs and governance through pooling, and which collectively bargained industries see their contribution base strengthened or weakened.

For single-employer plans, that means a present-tense governance problem and a future workforce-and-balance-sheet problem. For multiple-employer plans, it means an opportunity to use scale more intelligently, if providers can show that they deserve the trust such scale requires. For multiemployer plans, it means AI isn’t merely a technology story. It is a funding story in formation.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law, Bloomberg Tax, and Bloomberg Government, or its owners.

Read more in this series.

Author Information

Samuel W. Krause is a partner at Hall Benefits Law.

Write for Us: Author Guidelines