Return-to-office arguments often lean on mentoring. The claim is tidy: bring people back, and institutional osmosis will resume. The problem is that the research most worth trusting does not promise big payoffs from mentoring, and the demographic math says the real use case is narrower and more urgent. Mentoring works modestly, and it works best when it is built like training for specific skills across generations, not as a cure-all for culture.
What the evidence actually shows
Two large syntheses set the boundaries. A meta-analysis in the Journal of Applied Psychology found that mentored protégés do better on average than their non-mentored peers, but the effect sizes for hard career outcomes are small. Associations with compensation were near r ≈ .12, with promotions around r ≈ .31, while subjective outcomes such as career satisfaction were larger (Allen, Eby, Poteet, Lentz and Lima, 2004). That is a case for mentoring with realistic expectations, not for magical thinking.
A broader meta-analysis across youth, academic, and workplace contexts reached a similar conclusion. Mentoring is associated with favorable outcomes across behavioral, attitudinal, health, relational, motivational, and career domains, yet the effects are generally small. For example, corrected correlations were about rc = .14 for career attitudes and rc = .16 for situational satisfaction. Workplace mentoring sits in the middle of the pack, with typical effects in the .03 to .19 range, while academic mentoring tends to be stronger (Eby, Allen, Evans, Ng and DuBois, 2008). The authors caution against overselling mentoring given these magnitudes. They also aggregate across mentoring types, which means popular formats such as “reverse mentoring” are not isolated in a way that lets us claim specific causal effects for them.
A plain reading follows. Mentoring is a sound practice if you define success in terms of small improvements in satisfaction, attitudes, and developmental behaviors, and if you accept that objective career gains will be incremental. The meta-analytic record does not say anything about whether those benefits hinge on being physically co-located. It does say that programs should be designed and measured with the likely effect sizes in mind.
The real problem to solve: the age-skills gap
The demographic clock is not theoretical. The OECD reports that the old-age dependency ratio in member countries rose from 19 percent in 1980 to 31 percent in 2023, and is projected to reach 52 percent by 2060. Without productivity gains or policy changes, per capita GDP growth would slow by roughly 40 percent, from 1.0 percent a year in 2006–19 to 0.6 percent in 2024–60 (OECD, 2025). Only a third of adults aged 60 to 65 participated in training in 2023, compared with more than half of those aged 25 to 44 (OECD, 2025). That is a structural skills problem.
Viewed through this lens, mentoring is not a perk or a culture lever. It is a pragmatic tool for two concrete jobs. First, transfer institutional and customer knowledge from experienced workers before it walks out the door. Second, help experienced workers acquire new skills fast enough to stay productive, especially where formal training routes are thin for late-career employees. The first flow moves from older to younger. The second must flow back. Anything else wastes scarce time.
This is also where leaders should be honest about what mentoring can and cannot do. The meta-analyses suggest you are unlikely to see large, near-term jumps in pay or promotion probability driven by mentoring alone (Allen et al., 2004; Eby et al., 2008). You can, however, expect improvements in attitudes and satisfaction that support retention and the day-to-day cooperation that makes knowledge handoffs stick. That is exactly what an aging workforce needs.
Build mentoring like operations, not folklore
If mentoring is meant to close age-skills gaps, treat it like training design with guardrails that match the evidence.
- Define narrow outcomes. Track measures the research shows are responsive, such as career attitudes and situational satisfaction, alongside task-level proxies of skill transfer, like the number of systems a late-career employee can now operate independently or the cycle time for a handover. Expect small effect sizes and plan sample sizes and timelines accordingly (Allen et al., 2004; Eby et al., 2008).
- Make it bidirectional by default. Pairings should have two explicit learning goals, one for institutional knowledge moving to newer staff and one for tools or processes moving to experienced staff. Reverse mentoring is popular rhetoric, but rigorous evidence on its distinct impact is limited in the syntheses, so design it as one channel in a balanced exchange rather than the hero format (Eby et al., 2008).
- Time-box and scale. Use 8 to 12 week sprints with clear skill targets and documented artifacts. Rotate cohorts quarterly to reduce dependency on any single relationship and to spread exposure. The meta-analytic pattern that academic contexts yield stronger effects suggests that structure and curriculum help, so borrow that discipline even in workplace programs (Eby et al., 2008).
- Train the mentors. Provide short primers on feedback, goal-setting, and psychological safety. You do not need a sprawling curriculum to move the needle on attitudes and satisfaction, but you do need a shared playbook so that pairs know what “good” looks like (Allen et al., 2004).
- Measure participation asymmetries. Given that only a third of 60 to 65 year-olds engage in training, do not assume equal uptake. Make it easy for experienced workers to opt in with protected time, simple sign-up, and visible executive support. Monitor who participates and who benefits so you can close gaps rather than entrench them (OECD, 2025).
- Report results in language a board can hear. Anchor expectations around modest, leading indicators that accumulate over time, such as improved satisfaction, smoother handovers, and reduced rework. Do not claim that mentoring will move compensation or promotions materially in the next quarter. The best meta-analytic evidence says that is not how this variable behaves (Allen et al., 2004; Eby et al., 2008).
None of this requires a universal return-to-office policy. The evidence base does not tie mentoring’s modest benefits to a particular location. What it does tie them to is the presence of mentoring at all, and to the use of structure and goals that fit the outcomes mentoring can realistically move. If leaders want to bring people together for mentorship on specific days or for specific cohorts, they should say so directly and measure whether those sessions produce the desired skill-transfer artifacts.
The practical read
Stop using mentoring as a catch-all justification for sweeping office mandates. The stronger case, and the one the numbers support, is humbler and sharper. Use mentoring to do two things at scale: capture hard-won knowledge from people nearing retirement, and help those same people keep pace where formal training lags. Design for the outcomes the research says are most likely to shift, and hold yourself to reporting those movements plainly. In an aging economy with a training participation gap among older workers, that is not a nice-to-have. It is the work.
Sources
- Allen, Eby, Poteet, Lentz and Lima (2004). Career benefits associated with mentoring for protégés: a meta-analysis. Finds positive but small associations with objective outcomes and larger effects on subjective outcomes, underscoring realistic expectations. https://www.semanticscholar.org/paper/Career-benefits-associated-with-mentoring-for-a-Allen-Eby/a13c80a7f263f4378c07b474a0628105b91c6d28
- Eby, Allen, Evans, Ng and DuBois (2008). Does Mentoring Matter? A Multidisciplinary Meta-Analysis. Synthesizes 112 studies, reports generally small effects across domains and stronger results in academic settings, cautioning against overselling mentoring. https://pmc.ncbi.nlm.nih.gov/articles/PMC2352144/
- OECD (2025). OECD Employment Outlook 2025. Documents the rising old-age dependency ratio and the lower training participation among 60–65 year-olds, framing the age-skills gap mentoring should target. https://www.oecd.org/en/publications/oecd-employment-outlook-2025_194a947b-en.html