AI Exposure Hits 24% of Global Jobs, BofA

Bank of America Global Research published a note on May 2, 2026 estimating that roughly 24% of jobs worldwide could be "exposed" to artificial intelligence-driven automation. That headline figure, described by BofA as "almost a quarter" of global roles, immediately reframes the investment case for labour‑intensive sectors and for companies supplying AI infrastructure and services. Investors and corporate boards should treat the estimate as a directional shock rather than a deterministic forecast: exposure does not equal instantaneous displacement, but it signals material reallocation risk across occupations. This article places the BofA estimate in historical and cross‑agency context, presents a data deep dive of what "exposure" means quantitatively, and draws out plausible sectoral winners and losers while noting the policy and transition risks that could alter outcomes.

Context

The BofA finding arrives in a period of rapid AI deployment across enterprise IT stacks, customer service and creative workflows. Over the past three years cloud providers and chipmakers have accelerated capital spending: public filings show hyperscaler capex growth of double digits year‑over‑year in several quarters since 2023, supporting faster AI model training and deployment cycles. At the same time, labour markets in developed economies are structurally tight; employers face elevated hiring costs and wage pressures, which can make automation investments more economically attractive despite high upfront capex. The interaction of narrower labour supply, cheaper compute per model inference, and improved AI capabilities is a key backdrop for why a 24% exposure estimate is taken seriously by institutional investors.

BofA characterises "exposure" as the share of roles where a substantial share of tasks could be automated or materially augmented by current‑or‑near‑term AI capabilities, not the share of jobs that will vanish overnight. Historical precedents—such as the mechanisation of agriculture in the 20th century—show that task automation tends to change job content more often than it eliminates entire occupations in a single step. Nonetheless, transitional effects can be rapid for narrow occupational cohorts: clerical roles, repetitive back‑office functions, and certain legal and accounting tasks can be redesigned or consolidated within a short window once firms standardise AI tools. For fixed‑income and corporate credit analysts, this matters because EBITDA profiles and labour cost trajectories may diverge from consensus as firms reinvest savings or redistribute headcount.

Data Deep Dive

BofA's May 2, 2026 research note provides the 24% figure as its central estimate; the bank did not claim 24% equals near‑term unemployment but used it to flag potential exposure across job taxonomies. That 24% compares with other major studies: the OECD's 2019 report estimated about 14% of jobs at high risk of automation and an additional 32% likely to face significant change, while McKinsey's 2017 analysis forecast that up to roughly 30% of hours worked could be automated by 2030 under their midpoint scenario. The differences in these figures reflect methodology—OECD focused on task routineness and cross‑country labour regulation, McKinsey modelled hours and economic substitution, and BofA framed exposure specifically in the context of contemporary, generative AI capabilities (BofA, May 2, 2026).

To translate exposure into scale: if one accepts a 24% exposure rate against a working population of approximately 3.0–3.5 billion employed people globally (ILO baseline ranges vary by vintage), the arithmetic implication is a cohort in the hundreds of millions of roles where task content and labour value could be reallocated. The precise tally depends on country‑level labour intensity in services and manufacturing: advanced economies with larger shares of white‑collar employment may see higher proportional exposure, while lower‑wage emerging markets with larger shares of informal and manual labour may see different patterns of automation adoption. BofA's note makes explicit that the timing, depth and geographic concentration of job exposure will determine macroeconomic and corporate outcomes.

Sector Implications

Technology vendors and AI infrastructure providers are the most direct beneficiaries in BofA's scenario: firms that own cloud capacity, application‑layer AI, and accelerators are positioned to capture recurring revenue from model hosting, inference and toolchains. Market leaders with integrated cloud and enterprise software suites—alongside chipmakers providing high‑performance processors—stand to expand margins if adoption accelerates and per‑user revenue grows. This is why equities such as MSFT and NVDA (among others) are often singled out in market commentary as being bid to the extent AI spend decouples from general IT budgets, even while cyclical software and services businesses face margin pressure if they must heavily subsidise migration for large corporate clients.

Conversely, firms with large back‑office payrolls and thin margins will face two levers: invest to automate and protect margins, or retain labour and accept rising operating costs. Financials, professional services and traditional retail are examples where business models will bifurcate: those that invest in AI may compress headcount growth but expand scale and new product lines, while those that do not invest risk margin erosion. For corporate credit investors this bifurcation increases idiosyncratic risk; two similarly rated companies in the same industry could diverge sharply in credit metrics depending on adoption strategy and execution.

Risk Assessment

Transition risk is the primary near‑term hazard. Rapid automation in a concentrated set of occupations can lead to demand shocks, wage stagnation in affected cohorts, and political pushback that materialises as regulation or labour protections. Policymakers in major economies are already debating disclosure requirements for AI use, worker retraining programmes, and taxation of digital labour substitutes; any material policy intervention could alter the investment case for capital expenditure on AI. For sovereign debt and municipal credits, an accelerated decline in payroll tax receipts in exposed sectors could create revenue pressures at the local level, particularly where economies are specialised in affected industries.

Second, there are adoption and implementation risks: integrating AI into regulated workflows—healthcare, legal practice, and certain financial services—requires governance, model validation and human oversight. These compliance costs can blunt the ROI on automation in the short term. Third, the distributional impact—wage polarisation and possible aggregate demand effects—poses macro risk to cyclical consumption if a nontrivial share of middle‑income earners face prolonged income disruption. Stress testing and scenario analysis using BofA's 24% exposure as an adverse case therefore become higher‑priority tasks for institutional risk teams.

Fazen Markets Perspective

Fazen Markets interprets the BofA estimate as a clarifying datapoint that accelerates an existing investment theme rather than creating a wholly new one. Our contrarian view is that markets may currently underprice the dual effect of (1) accelerated productivity and (2) concentrated distribution of gains to capital owners in the near term. Historically, technological revolutions have increased aggregate output while producing short‑to‑medium term dislocations; the critical variable is the speed of labour reallocation and the effectiveness of institutional responses. If AI adoption is front‑loaded by large incumbents funding capex with healthy balance sheets, equity returns for select AI infrastructure names could outpace the broader market even as aggregate employment measurements adjust more slowly.

This perspective suggests investors should emphasise granular, bottom‑up analysis: assess company‑level exposure to labour substitution, track capex intensity and operating leverage in filings, and model potential regulatory scenarios. We also highlight the asymmetric information play: firms that transparently disclose process automation plans and retraining programmes may face less execution uncertainty and reputational risk. For further research on technology adoption and macro linkages see Fazen Markets tech coverage and our labour market primer on structural shifts labour markets.

Frequently Asked Questions

Q: How fast could job displacement occur under the BofA scenario?

A: Timelines vary by occupation; BofA frames exposure over a multi‑year horizon rather than immediate elimination. Empirical adoption often follows an S‑curve: pilot and consolidation phases can take 1–3 years in enterprise settings, with broader rollout taking an additional 3–7 years depending on regulatory friction and incumbent investment cycles. High‑volume clerical and customer‑facing tasks are most susceptible to faster transitions given lower integration complexity.

Q: Are there historical precedents that inform likely outcomes?

A: Yes. The mechanisation of agriculture and the diffusion of information technology in the 1990s both led to productivity gains and net job creation over multi‑decadal horizons, but they also produced sectoral and regional dislocations that lasted a generation in certain localities. The key differences today are the speed of software diffusion, lower marginal cost of scaled-up AI services, and the concentration of capabilities in a handful of large platforms—factors that amplify both productivity upside and distributional risk.

Bottom Line

BofA's May 2, 2026 24% exposure estimate is a material input for scenario analysis: it increases the urgency for investors to stress‑test portfolios for labour‑substitution risk and to identify concentrated winners in AI infrastructure. Institutional investors should prioritise granular, issuer‑level assessments and policy‑sensitivity modelling rather than rely on headline macro narratives.

Disclaimer: This article is for informational purposes only and does not constitute investment advice.