Zimbabwean banks find AI can broaden access—especially anomaly detection, which explains most measured differences—but rollout is constrained by digital illiteracy, poor internet and high costs, and customers remain wary of opaque credit-scoring and chatbot decisions.

Main Finding

AI adoption in Zimbabwean commercial banks—particularly anomaly detection systems, AI credit scoring, chatbots, and predictive analytics—can materially advance financial inclusion by improving access, speeding decisions, and strengthening security. However, benefits are constrained by limited digital literacy, poor internet connectivity, high application costs, and a rural–urban divide; anomaly detection had the largest measured impact (explaining 62.3% of observed outcome differences).

Citation: Muswaburi, J. T., Mandava, P., Mpondwe, N., & Mumanyi, M. (2026). The Impact of Artificial Intelligence on Financial Inclusion in Zimbabwe's Banking Sector. International Journal of Latest Technology in Engineering, Management & Applied Science (IJLTEMAS). DOI: 10.51583/IJLTEMAS.2026.15020000087

Key Points

• Study design: mixed methods (quantitative survey + qualitative interviews) focused on AI use-cases in banking: AI credit scoring, chatbots/virtual assistants, anomaly/fraud detection, and predictive analytics/robo-advisors.
• Sample: 293 survey respondents (stratified random sampling from employees of major commercial banks in Harare) and 12 semi-structured interviews with management stakeholders.
• Perceptions (selected means, 1–5 scale):
  • Predictive analytics aids decision-making: 4.14
  • AI recommendations improve budgeting: 4.05
  • AI expands access for unbanked: 3.91
  • AI enhances online safety / helps detect suspicious activity: ~3.78–3.91
  • AI credit decisions are transparent: 3.25 (lowest)
  • Regular use of chatbots: 3.45 (lower adoption)
  • Major perceived barriers (high mean values): limited digital literacy (4.21), poor internet (4.13), high costs (4.07), urban–rural disparity (4.10)
• Hypotheses tested: H1 (AI credit scoring → access), H2 (chatbots → usage), H3 (anomaly detection → security), H4 (predictive analytics/robo-advisors → financial literacy). Results emphasize strong institutional support for decision-making and fraud detection, user concerns about transparency and operation of credit assessments/chatbots, and empirical support for anomaly detection’s outsized effect (62.3% explained variance).
• Governance and fairness concerns: algorithmic bias, lack of transparency and recourse, weak data governance and privacy risks were identified as critical constraints that could reverse inclusion gains if unaddressed.
• Recommendations from authors: expand accessible AI decision tools, improve digital literacy programs (with regulatory support), and allocate resources for underserved/rural communities and infrastructure.

Data & Methods

• Philosophy/approach: Pragmatism with an abductive reasoning approach.
• Design: Mixed-methods — correlational quantitative analysis complemented by exploratory qualitative interviews.
• Quantitative:
  • Population: 1,754 bank employees in Harare (923 management, 831 regular staff).
  • Sample: 293 (calculated via Raosoft), stratified random sampling.
  • Tools: Online questionnaires; descriptive statistics; multiple regression analysis in SPSS.
  • Reliability: Pilot testing and Cronbach’s alpha > 0.7 for constructs.
• Qualitative:
  • Sample: 12 purposively selected management interviewees (face-to-face and telephone).
  • Analysis: Thematic coding in NVivo; triangulation and member checking used to bolster trustworthiness.
• Ethics: Removal of personal identifiers; informed consent and participant rights communicated.
• Limitations noted or implied: cross-sectional, employee-focused sample centered in Harare (not direct customer panels across rural areas), reliance on perceptions/self-report, potential selection/coverage bias from online survey distribution.

Implications for AI Economics

• Market expansion & credit supply: AI credit scoring using alternative data can lower information frictions and expand credit to previously “thin-file” borrowers, potentially increasing loanable funds and financial market participation at lower marginal screening costs.
• Risk pricing and lender behavior: Better predictive models can reduce adverse selection and moral hazard costs, enabling lenders to offer smaller, cheaper loans to underserved segments—however, model quality and bias determine whether risk-adjusted pricing improves access or codifies exclusion.
• Security externalities and trust: High-performing anomaly detection raises expected returns to digital banking by reducing fraud costs and increasing customer trust; this can accelerate adoption and network effects, especially in mobile-money-heavy economies.
• Distributional effects & inequality: Gains are likely to be concentrated among digitally literate and connected urban users unless targeted interventions (digital literacy, subsidized connectivity, local-language interfaces) are implemented—AI can thus both reduce and exacerbate inequality depending on complementary investments and policies.
• Regulatory & governance economics: Effective data governance, transparency mandates, auditability, and consumer recourse mechanisms are economic complements that shape adoption incentives, competition, and welfare outcomes. Without them, algorithmic bias and opacity can create welfare losses and market failures.
• Investment priorities: From a policy-economic perspective, returns to investing in broadband, affordable applications, and digital skills may be large because they multiply the impact of AI tools on inclusion; subsidies or public–private partnerships targeting rural infrastructure could yield high social returns.
• Research & evaluation needs: Economists should prioritize causal evaluation (randomized or quasi-experimental designs) of AI interventions on actual consumer outcomes (account uptake, usage, credit uptake/defaults, welfare), disaggregated by rural/urban, gender, and income to measure distributional impacts and detect harmful biases.
• Policy recommendations (actionable):
  • Mandate transparency and explanations for automated credit decisions; require independent audits for fairness and accuracy.
  • Invest in digital literacy programs targeted at underserved groups.
  • Support affordable connectivity and subsidized on-ramps (e.g., shared digital kiosks, community agents) in rural areas.
  • Encourage banks/regulators to adopt data protection, consent, and redress frameworks alongside AI deployment.
  • Promote pilot programs with rigorous impact evaluation before large-scale rollout.

Limitations and caveat: Findings are largely perception- and employee-based with cross-sectional data from Harare banks; causal claims about household-level financial inclusion outcomes require customer-level, longitudinal, or experimental evidence.