On March 19, 2026, SME Banking Club hosted a deep‑dive workshop exploring one of the most urgent and fast‑moving topics in financial services today: the rise of AI agents in SME banking and lending. The event brought together industry practitioners, technology innovators, and SME finance experts to discuss what “agentic AI” really means, how banks are already applying it, and what the near future will look like.
The workshop featured an impressive lineup of speakers and practitioners—including Olena Gryniuk, Daniel Huszar, Christian Ruehmer, Tamás Csontos —all sharing real‑world applications, demos, and strategic insights on the adoption of AI agents across CEE.
AI Adoption in SME Banking: Where the Industry Stands
Olena Gryniuk opened the session by presenting SME Banking Club’s latest regional insights. A study conducted in late 2025 showed:
- 66% of financial institutions in CEE have already started their AI journey (pilots or live use cases).
- 15% use AI broadly in daily operations.
- 19% do not use AI at all—although this figure is expected to shrink rapidly.
The strongest adoption areas so far include:
- Risk & credit decisioning (52%)
- Operations automation (43%)
- Customer service & digital experience (38%)
Banks like BNP Paribas, Alior Bank, ING Poland, and Santander Bank are already implementing AI assistants, automated document search, branch avatars, and internal procedure‑search agents.
From ChatGPT to Enterprise Agents: What’s the Difference?
Technology consultant Daniel Huszar demystified the term AI agent, stressing that enterprises must combine:
- Deterministic software (rules, calculations, workflows)
- Non‑deterministic AI models (LLMs like GPT‑4/5)
This hybrid architecture matters because financial institutions operate in a regulated, auditable, and risk‑sensitive environment.
“An AI agent is not just ChatGPT. It’s a structured, repeatable workflow with instructions, memory, constraints, and integrations.”
Daniel demonstrated how even simple no‑code agents can transform messy, unstructured SME lending data into actionable, structured briefs—saving hours of manual sifting through emails, spreadsheets, and documents. These personal‑level agents also help business and tech teams collaborate better by aligning on workflows and expectations before building enterprise‑grade solutions.
The Three Pillars Needed Before Implementing AI in SME Finance
Christian Ruehmer, CEO of Q-Lana, emphasized that successful AI adoption requires strong foundations in three domains:
- Data Management
Banks must move from information to intelligence through:
- disciplined data collection
- eight core data domains (customers, relationships, financials, tax, credit process, behaviour, collateral, policies)
- clean identifiers, metadata, and governance
- minimum viable data architecture
Without data discipline—not volume—AI will fail.
- Risk Management
AI agents depend on precise and consistent risk logic:
- PD, LGD, EAD
- Expected vs. unexpected loss
- Regulatory‑aligned formulas
- Scenario planning
- Clear risk appetite frameworks
This enables a fundamental shift:
from approving loans after clients apply → to offering products already matched to pre‑approved risk appetite.
- Customer Centricity
Banks must shift from selling products to solving customer problems:
- “jobs‑to‑be‑done” libraries
- cash‑flow‑aligned financing
- diagnostic conversations
- cross‑functional “solution pods”
- ongoing monitoring & coaching
AI agents can guide relationship managers with scripts, insights, and analysis—elevating rather than replacing human expertise.
Live Demo: How Enterprise AI Agents Actually Work
Tamás Csontos from Intuitech showcased a live end‑to‑end demo of an SME loan origination AI agent already running in production at banks such as Erste BCR Romania.
The agent performs five core functions:
- Document ingestion
Reads any format—PDF, scans, XLS, images, even handwritten—structured or unstructured.
- Data extraction, validation, and cross‑checking
Detects:
- missing documents
- inconsistent values
- incorrect formats
- mismatches across files
Using their own “bounding box” feature, the agent highlights the exact field in the document where an issue occurred.
- Expert-level analysis
Performs:
- DSCR calculations
- market analysis
- financial trends
- risk factor evaluation
- policy checks
All fully auditable with formulas and timestamps.
- Automated communication
Drafts client emails requesting corrections, additional documents, or explanations—in the bank’s tone of voice.
- Preparing decision-ready credit memos
Assembles the entire credit memo using:
- validated data
- rule outcomes
- analyst-level insights
- bank’s standard format
RM or risk manager may edit before submission.
The Results: Efficiency, Accuracy, and Time-to-Yes
Banks using these solutions report:
- 60–70% reduction in loan processing time
- Reduction of cycles from 7 weeks → a few days or hours
- 95%+ accuracy, contractually guaranteed
- Human accuracy averaging 80–85% in comparison
- Potential 15–20% increase in loan volumes thanks to market-leading time-to-cash
- Freeing experts to focus on judgment, relationships, and revenue—not admin work
As Tamás noted:
“We don’t replace risk officers—we remove the drudgery so they can focus on what humans do best.”
Key Discussion Themes from the Expert Q&A
The workshop closed with a rich discussion covering:
Data quality challenges
How to know when data is “good enough” for AI, and how to train the system using synthetic data when historical datasets are poor.
ROI and TCO expectations
Banks want measurable outcomes—not innovation theater.
AI agents show quantifiable gains, but also strategic, less tangible benefits (customer satisfaction, market share, speed).
Impact on junior talent
Will fewer manual tasks make it harder to train future risk managers?
Consensus: AI changes—but doesn’t eliminate—the learning path.
Young analysts must learn:
- how to review AI output
- how to ask the right questions
- how to understand risk fundamentals
AI in ongoing monitoring
Huge opportunity in:
- covenant checks
- invoice monitoring
- early warning systems
- behavioural analytics
Not just origination.
Conclusion: AI Agents Are Becoming the New Operating System of SME Banking
The workshop made one thing clear:
AI agents are no longer a futuristic idea—they are already reshaping SME lending and banking operations across Europe.
Banks that prepare their data, risk frameworks, and customer-centric models today will be the ones able to leverage AI at scale tomorrow.
The shift is as much organizational and cultural as it is technological.
The winners will be institutions that embrace AI not as a tool, but as a transformation of how banking gets done.
If you’re looking to transform your SME lending process today, reach out to our team to learn how they can support and accelerate that transformation.
