Windfall Fellowship Program
Welfare in the AI Transition: Brazil Case Study
Using this content? Please see our Content Permissions policy.
This case study examines whether Brazil’s welfare system can protect workers from AI-related disruption, particularly in clerical, administrative and other office-intensive roles. We find that Brazil can identify and pay households at population scale through CadÚnico, Pix and Bolsa Família, but lacks an already established system linking displaced workers to income support and retraining. Oversight and appeals have also failed to keep pace with automated INSS benefit decisions. We identify reforms Brazil can make now, alongside the institutional and fiscal changes needed for a more durable response.
Published August 18th 2026
Welfare Resilience Snapshot · Brazil
Risks
What pressures does the welfare system face?
Policy levers
What instruments can respond to these pressures?
Enablers
Can those instruments actually function effectively?
1. OVERALL RISK EXPOSURE
Nature of risk
Mostly gradual
Most workers earn in informal, self-employed or casual activities where earnings fluctuate frequently. AI adoption is likely to amplify these fluctuations.
TWO URGENT OPERATIONAL PRESSURES
Displacement in clerical and administrative roles; the banking sector showed a clear signal.
Over-automated welfare administration inside the INSS.
Dominant channels
SHORT TERM
Disruptive
Earnings shocks arrive quickly: food price changes, droughts, shifts in platform demand.
MEDIUM TERM
Transitional
Risks build gradually as tasks change in outsourcing, logistics and digital work.
LONG TERM
Structural
Tied to persistent informality and limited redistribution capacity.
MORE SLOWLY
Where risk is concentrated
The missing middle remains the sharpest exposure
Poorest households - Bolsa Família protects at the bottom of the distribution.
The missing middle - Informal workers and displaced formal workers above the cash-transfer threshold still fall between the two.
Formal workers - Seguro Desemprego cushions briefly after job loss.
2. BINDING CONSTRAINTS
Weak transition architecture for displaced workers
No standing transition system
Training institutions, labour-market data and some sectoral retraining capacity exist, but nothing links income support, worker identification and retraining into a forward-looking system.
Over-automation without strong safeguards
INSS automated benefit analysis
The INSS has expanded automated benefit analysis faster than audit visibility, human review and appeals protection have kept pace.
Financing and regional gains weakly connected
No dedicated welfare-financing system
Tax reform and dividend taxation begin to broaden the revenue base, but there is no welfare-financing channel and no mechanism turning AI-linked regional growth into broader territorial gains.
3. CAPACITY PROFILE
DISRUPTIVE · SHORT TERM
Coping capacity
IN PLACE
Strong delivery rails and a large social registry
Proven emergency cash capacity
LIMITS
Normal rules still exclude the missing middle by design
TRANSITIONAL · MEDIUM TERM
Adaptive capacity
IN PLACE
The banking sector's collective agreement provides a retraining model
LIMITS
No clear mid-career retraining and reallocation pathway for displaced workers outside that agreement
STRUCTURAL · LONG TERM
Transformative capacity
IN PLACE
VAT reform and the 2025 dividend-tax change begin to diversify the revenue base
LIMITS
No welfare earmark yet
No broader replacement for payroll-linked financing
ACCESS & DELIVERY · CROSS-CUTTING
Systemic capacity
IN PLACE
Strong digital-state capacity
LIMITS
Oversight, transparency and appeals have not kept pace with automated decision-making
Core challenge
An institutional mismatch: digital capacity, fiscal change and policy ambition are advancing faster than the systems that connect them to fair protection in practice.
4. WHAT CAN BE DONE NOW VS LATER
Achievable now
Within existing law, institutions and fiscal capacity
External audit of automated benefit analysis
Bring automated benefit analysis back into routine external audit and public reporting.
Binding appeal-time targets
Set binding appeal-time targets and publish CRPS performance data.
Offline intake and identity verification
Strengthen offline pathways for claimants who cannot reliably navigate digital channels.
Extend the banking retraining model
Extend it to other high-exposure sectors through collective bargaining, using PBIA and Sistema S capacity to build AI-complementary curricula.
AS
ENABLERS
STRENGHTEN
Feasible only if enablers are strengthened
Blocked on legal, institutional or fiscal reform
Proactive RAIS-CadÚnico targeting system
Depends on: Inter-ministerial protocol, Legal basis for prospective use of identified records, Funding for follow-through
Standing income bridge for the missing middle
Depends on: Legislative expansion of existing instruments or a new automatic top-up mechanism
Enforceable AI rights in employment and welfare
Depends on: Passage and implementation of PL 2338/2023 or an equivalent legal framework, Institutional capacity to enforce it
Reduced payroll dependence in welfare financing
Or creating welfare-oriented value capture from data-center expansion.
Depends on: Additional political choices beyond the reforms already in place
Brazil’s most pressing welfare risk from AI is transitional. Banking provides the clearest realised example of employment pressure so far, while exposure extends more widely across routine clerical, administrative and office-intensive work. Women are particularly exposed because they are more concentrated in these occupations. Platform work and regional inequality may also leave displaced workers with fewer protected alternatives.
Applying Windfall’s Welfare Resilience Assessment Framework, this case study finds that coping capacity is Brazil’s strongest dimension. CadÚnico, Pix and Bolsa Família provide population-scale systems for identifying households, making payments and supporting incomes. Auxílio Emergencial showed that this infrastructure can deliver emergency cash at very large scale when political authorisation is in place.
Brazil’s weakest dimension is adaptive capacity. There is no standing system connecting worker identification, temporary income support and mid-career retraining. Informal workers and displaced formal workers above the Bolsa Família threshold can fall into a “missing middle” once short-term support ends. The banking sector’s collective agreement provides a model for retraining, but no comparable pathway exists across most exposed sectors. At the same time, automated INSS decisions have expanded without equivalent improvements in external audit, transparency, human review or timely appeals.
The report recommends beginning with governance reforms that can be delivered through existing institutions: routine external audit of automated INSS decisions, publication of denial rates, binding appeal-time targets and reliable in-person intake and identity-verification routes. The medium-term priority is to connect RAIS, CadÚnico and Sistema S so that displaced workers can be identified and directed towards income support and retraining. A standing income bridge for the missing middle, enforceable rights around high-stakes automated decisions and a broader funding base for welfare will require further legal and political choices.
Brazil already has much of the delivery machinery it needs. Its preparedness now depends on connecting those systems, strengthening safeguards and creating a transition pathway before labour market disruption becomes widespread.
See also
A Framework for Welfare Resilience in the Age of AI
Read more
