Lending institutions across Central and Eastern Europe, South‑Eastern Europe and the Baltic states operate at the intersection of fragmented data and intensifying risk. Recent risk assessments show that Stage 2 exposures – loans whose credit risk has significantly increased – reached EUR 1.6 trillion, or 9.4 % of total amortised cost loans by mid‑2025 and remained high at 9.2 % in preliminary third‑quarter data. These volumes are concentrated in commercial real estate and SME portfolios. Risk models built on inconsistent or delayed data may either reject viable borrowers or under‑price emerging risks, forcing banks to increase provisioning and eroding capital buffers.
Elevated Stage 2 balances have a direct impact on lending capacity. Provisioning costs rise, capital becomes tied up and appetite for new lending shrinks. The EIB’s CESEE Banking Lending Survey (H1 2025) reports that demand for credit across the region remains strong while supply has been weak since 2022, although banks expect a modest improvement. Two‑thirds of cross‑border banking groups intend to expand selectively in the region, signalling opportunity, but they must navigate uneven growth, elevated risk and tightened standards. Without better risk visibility, lenders can miss genuine opportunity and misallocate scarce capital.
The opportunity lies in harnessing artificial intelligence and modern data platforms to derive insight from messy, heterogeneous sources. AI‑driven analytics can synthesise signals from multiple jurisdictions, update risk views in near real time and surface early warning indicators that enable proactive intervention. By embedding these capabilities throughout the credit lifecycle – from origination to servicing – banks can turn fragmented data into a competitive advantage. This article outlines how AI transforms credit insight in the CEE/SEE/Baltic context, drawing on recent regulatory findings, market surveys and technology trends, and explains how Axe Finance’s Axe Credit Portal (ACP) operationalises these innovations.
Elevated risk and the search for resilience
Asset quality across Europe remains resilient by historical standards, yet the growing stock of Stage 2 loans signals pressure ahead. The European Banking Authority’s autumn 2025 risk assessment notes that Stage 2 exposures have risen by around EUR 240 billion since 2021 due to deterioration in commercial real estate, rising interest rates and geopolitical tensions. Commercial real estate loans show Stage 2 ratios of 17.1 % and SME portfolios 14.9 %, far above the overall average. While cost of risk remains low (around 48 basis points), the share of underperforming loans demands early action before losses crystallise.
The region’s macro backdrop is equally complex. Raiffeisen Research reports that CEE and SEE banks generated returns on equity of 15–20 % in 2024 – around eight percentage points above the euro‑area average – thanks to cautious monetary easing, solid economic growth and low unemployment. Profit pools reached roughly EUR 30 billion with Romania and Serbia achieving 20 % returns. Yet the same report warns that extra banking levies affect around 85 % of regional assets and that exposures to Russia have been reduced to around 4 % as groups refocus on EU markets. High profitability masks underlying volatility and the need for disciplined risk management.
Diverging macro scenarios underscore the importance of resilience. Deloitte’s 2026 banking outlook highlights that persistent inflation, tariff‑driven price pressures and diverging consumer sentiment could test revenues and profitability. In its baseline scenario, real GDP growth in 2026 could slow to around 1.4 %, consumer sentiment moderates and the labour market softens. Against that backdrop, banks must defend margins, diversify fee income and prepare for competition from non‑banks. Stablecoins and tokenised deposits may also disrupt payment rails, forcing decisions on whether to issue, custody or process them.
Data fragmentation and cross‑border complexity
CEE and SEE banking groups often span multiple jurisdictions, each with its own supervisory requirements, data registries and accounting practices. Public registries vary in their coverage and timeliness; SME financials may arrive late or in inconsistent formats; and group structures obscure exposure concentrations. The EIB survey indicates that lenders generally rate market potential in the region as medium or high rather than low. To seize this potential, lenders must reconcile data across subsidiaries and service lines.
Digital channels partially mitigate fragmentation. The SME Banking Club’s 2025 survey of digital SME lending found that about 43 % of banks and non‑bank lenders in the region offer online or mobile loan applications, with roughly one in four providing fully digital processes. Yet this still leaves a majority of applications reliant on manual data capture and interpretation. The CEE, SEE & Baltics Summit 2025 emphasised that AI adoption remains limited – with only around 20 % of innovations fully adopted – due to cultural and regulatory reluctance, data transparency issues and concerns about cyber risks. Without harmonised data pipelines, AI cannot deliver its full value.
Turning data into insight: predictive analytics and behavioural scoring
AI’s first contribution is to make sense of partial information. Predictive analytics can infer repayment probabilities by learning how cash‑flow seasonality, invoice timing and client concentration affect default risk. Instead of waiting for full financial statements, models update in real time as new transactions or behavioural signals arrive. This is particularly valuable when borrowers lack deep credit histories or operate across multiple jurisdictions. Industry research stresses that banks must modernise data architectures and adopt generative AI to meet new demands.
Behavioural scoring complements traditional bureau data by incorporating patterns in account usage, payment behaviour and responsiveness during onboarding. By combining these signals, lenders reduce false negatives for “new to credit” customers and improve portfolio granularity. The European Central Bank’s 2025 Q3 bank lending survey shows how quickly credit standards can tighten in response to risk perceptions, as banks reported a slight net tightening in credit standards for firms and moderate tightening for consumer credit. An AI‑enabled behavioural model provides a more stable foundation during such swings.
Agentic AI – collections of autonomous software agents that collaborate to perform complex tasks – hints at the next stage of evolution. Deloitte observes that real‑world agentic AI applications are still uncommon due to regulatory challenges, model risk and legacy system integration, but it encourages banks to pursue high‑impact, low‑risk use cases. One promising area is continuous KYC maintenance, where different agents fetch public data, score risk and file regulatory reports using protocols like Model Context Protocol; Italian bank Intesa Sanpaolo is already piloting a multi‑agent system. For CEE lenders, such architectures could harmonise onboarding and monitoring across jurisdictions.
Early warning and proactive risk management
Machine‑learning‑driven early warning systems help lenders detect emerging distress before a loan becomes non‑performing. The EBA notes that cyber and ICT risks are among the top operational risks; distributed denial‑of‑service attacks accounted for a large share of reported incidents in 2025 and the Digital Operational Resilience Act (DORA) will harmonise incident reporting. While this concerns operational resilience, it highlights the importance of proactive monitoring and cross‑border coordination – principles that also apply to credit risk.
Early warning models track subtle shifts such as declining average balances, rising overdraft utilisation or delayed supplier payments. These signals trigger interventions: outreach to the customer, covenant review, or renegotiation before arrears accumulate. Combined with scenario analysis, early warning helps banks allocate capital more efficiently and reduces Stage 2 migration. The Vienna Initiative’s NPL Monitor underscores the role of macro drivers in non‑performing loans and the need for timely interventions (see the NPL Monitor H2 2024 report). AI allows banks to move from quarterly reviews to continuous monitoring across the portfolio.
Automating the full credit lifecycle with AI
To turn predictive insight into operational outcomes, AI must be embedded throughout the credit lifecycle. This means integrating risk models into origination, scoring, underwriting, servicing, collateral and collection processes. Banks must also ensure explainability and auditability to satisfy internal and external governance. Industry analysts argue that layering AI agents on top of existing robotic‑process‑automation frameworks can deliver measurable benefits without a full system replacement. Over time, banks may adopt an “agentic by design” architecture with microservice‑like applications. Either way, the goal is to connect analytics to workflow.
Axe Credit Portal demonstrates what end‑to‑end automation looks like in practice. ACP replaces manual, siloed processes with an integrated platform that automates everything from client onboarding and credit application to servicing and limit management. Its AI layer extracts data from documents via natural‑language interfaces, performs automated financial spreading and narrative generation, and generates risk scores with explainability. Eligibility rules, peer comparison and sentiment analysis enrich the decision process, while behavioural and transaction monitoring deliver early warning and cross‑sell recommendations. Automated affordability checks reconcile payslip and tax data with observed cash flows and simulate the impact of new obligations.
Because ACP spans the credit lifecycle, it enables standardised workflows across subsidiaries – a critical requirement for cross‑border groups such as OTP Group, a leading lender in CEE. OTP has implemented Axe Finance solutions across multiple subsidiaries, enhancing consistency and governance and allowing faster decision cycles. While specific metrics remain confidential, the qualitative benefits include shorter time‑to‑yes, reduced manual effort and improved audit trails, all of which support regulatory compliance and capital efficiency.
Conclusion
The CEE, SEE and Baltic banking markets are poised between opportunity and risk. High profitability and strong credit demand coexist with elevated Stage 2 exposures and macro uncertainty. Fragmented data and cross‑border complexity limit lenders’ ability to respond quickly. By embracing AI and modern data architectures, banks can turn these constraints into differentiators. Predictive analytics and behavioural scoring expand access to finance for thin‑file borrowers; early warning systems enable proactive risk management; and end‑to‑end platforms like ACP operationalise insight across the full credit lifecycle. The institutions that succeed will be those that invest in data unification, adopt AI responsibly and align technology with governance. In doing so, they will build resilience and seize growth in one of Europe’s most dynamic banking regions.
References
- European Banking Authority: Risk Assessment Report – Autumn 2025
- European Investment Bank: CESEE Banking Lending Survey 2025 H1
- Raiffeisen Research: CEE Banking Sector Report 2024
- Deloitte: Banking Industry Outlook 2026
- Deloitte: Agentic AI in Banking
- CEE, SEE & Baltics Summit 2025 – Key Takeaways
- European Central Bank: Euro area bank lending survey Q3 2025
- Vienna Initiative: NPL Monitor H2 2024
- Axe Finance: AI‑based lending solution