Tight credit markets present significant challenges for financial institutions, particularly those focused on B2B lending like SME banks, alternative lenders, and credit funds. Lenders face increased pressure to accurately assess risk, accelerate decision-making, and maintain profitability amidst economic uncertainty. For many, this means grappling with manual processes that slow down operations and fragmented systems that lack transparency [1].
Generative AI (GenAI) offers a transformative solution for financial institutions navigating these challenges. It can simultaneously increase accuracy, reduce operational costs, and accelerate decision-making in challenging market conditions. This isn't just theoretical; 98% of fund management firms have already invested in GenAI, indicating widespread adoption across the financial services sector [2].
This momentum suggests that automated lending solutions powered by GenAI are quickly becoming standard practice, offering a path to enhanced efficiency and accuracy. This article examines how GenAI is transforming loan underwriting in tight markets, offering financial institutions a pathway to enhanced efficiency, more accurate risk assessment, and sustainable growth.
GenAI is fundamentally changing how financial institutions approach loan underwriting. It represents a shift from purely human judgment to AI-augmented processes that can analyse vast datasets and identify patterns beyond traditional credit metrics [2].
The adoption of GenAI across the financial services sector is accelerating, driven by the promise of enhanced efficiency and cost savings [1]. Over 25% of financial institutions plan to deploy agentic AI solutions this year [1]. These autonomous systems are designed to make decisions and take actions towards a goal, and this high adoption rate demonstrates that automated lending solutions powered by GenAI are becoming the industry standard rather than the exception.
Financial institutions recognise GenAI's potential to transform their operations, particularly in data-intensive processes like underwriting where pattern recognition and predictive capabilities can significantly enhance decision-making accuracy.
GenAI is revolutionising credit risk assessment by analysing complex data patterns and identifying risk factors that traditional models might miss. This capability is particularly valuable in tight markets where accurate risk assessment can mean the difference between profitable lending and significant losses.
Banks are increasingly using AI for dynamic stress testing, liquidity risk modeling, and real-time risk analytics to make more informed lending decisions [3]. Dynamic stress testing, for instance, allows lenders to simulate various economic downturns to understand portfolio resilience, while real-time analytics provides continuous monitoring of risk exposures.
The adoption of AI for sophisticated risk analysis represents a significant advancement in how financial institutions assess creditworthiness. By implementing dynamic stress testing and real-time risk analytics, lenders can now evaluate loan applications with greater precision and adapt quickly to changing market conditions.
This trend is particularly relevant for SME lending, where traditional risk models often fall short due to limited data or unique business circumstances.
AI-driven analytics can uncover subtle patterns in financial behaviour that indicate creditworthiness beyond conventional metrics, enabling more accurate risk assessment and potentially expanding access to credit for previously underserved businesses. Beyond traditional data, modern platforms can integrate alternative data sources like transaction histories, utility payments, and even anonymised social media activity to build a more complete picture of an SME's financial health and stability [1].
Alternative credit funds, for example, are increasingly leveraging GenAI for enhanced predictive default modelling and early risk identification. By integrating vast datasets, including historical financial data and real-time market trends, GenAI models can identify subtle patterns and correlations that traditional models might overlook, allowing fund managers to anticipate defaults with greater precision.
This real-time risk assessment capability is crucial in the dynamic environment of alternative credit markets.
The effectiveness of GenAI in risk assessment fundamentally depends on data quality. Financial institutions often struggle with fragmented data sources, inconsistent formats, and incomplete records. Successful implementations address these challenges through dedicated data governance frameworks, automated validation processes, and continuous monitoring of data integrity across systems.
Leading platforms incorporate automated data validation, standardization, and enrichment capabilities that transform disparate, inconsistent inputs into structured, analysis-ready datasets—addressing a key challenge for institutions struggling with fragmented information systems [4]. This richer dataset, analysed by GenAI, allows for more nuanced risk profiling, which is crucial when navigating tight markets.
B2B lenders are implementing specific technical approaches to integrate diverse data sources. These include natural language processing (NLP) for extracting insights from unstructured documents, standardised API frameworks for connecting disparate systems, and automated data validation pipelines that ensure consistency across sources.
For SME lenders specifically, successful implementations often begin with targeted data integration projects focused on high-value information sources rather than attempting comprehensive integration simultaneously.
GenAI-powered automated underwriting systems are dramatically reducing loan processing times. This enables financial institutions to make faster decisions without compromising on risk assessment quality. This efficiency is creating competitive advantages for early adopters, particularly in markets where speed of approval can be a decisive factor for borrowers.
Consider the efficiency gains: automated workflows now account for less than 20% of SME loan approvals, highlighting the significant room for improvement through automation [1]. Darlington Building Society's implementation of the MSO loan origination platform, for instance, reduced application time by 10 days while enabling real-time application tracking for brokers [5].
The significant reduction in application processing time demonstrates the tangible efficiency gains that automated underwriting systems can deliver. By cutting 10 days from the application process, they improved operational efficiency and enhanced the customer experience.
This case study illustrates how technology-driven underwriting can transform a lengthy, paper-heavy process into a streamlined digital experience. The addition of real-time tracking further increases transparency and reduces uncertainty for all parties involved, addressing key operational bottlenecks.
Such improvements are particularly valuable in competitive markets where speed of service can be a key differentiator and high operational costs are a constant pressure point.
As financial institutions increasingly adopt GenAI for loan underwriting, they must navigate a complex and evolving regulatory landscape. Understanding and addressing these regulatory challenges is essential for successfully implementing automated lending solutions that remain compliant while delivering business benefits.
Nearly two-thirds of U.S. regulatory compliance professionals view technology-driven risk as the most significant market force likely to cause compliance issues for U.K. financial services firms in 2025 [6]. This statistic reveals a significant concern among compliance professionals regarding the integration of advanced technologies like GenAI into financial services operations.
The high percentage indicates that regulatory challenges are not peripheral but central to the implementation of AI in lending.
For financial institutions developing or deploying automated lending solutions, this highlights the need for robust compliance frameworks specifically designed for AI applications. Regulators are likely to increase scrutiny of AI-based decision-making systems, particularly in sensitive areas like credit underwriting where fairness, transparency, and accountability are paramount.
Institutions must therefore balance innovation with compliance to avoid regulatory penalties and reputational damage.
Addressing potential algorithmic bias is critical; institutions are increasingly using fairness audits, diverse data sets for training, and developing bias detection tools to ensure equitable outcomes [1]. Transparency through explainable AI (XAI) frameworks is also becoming essential for auditability [1, 6]. XAI refers to technologies that make AI decision processes interpretable to humans.
Beyond compliance, financial institutions implementing GenAI must prioritise robust cybersecurity frameworks. As automated lending systems process sensitive financial data, they become potential targets for cyber threats. Leading institutions are implementing end-to-end encryption, regular security audits, and advanced threat monitoring specifically designed for AI applications to mitigate these risks.
For SME financial institutions with potentially more limited compliance resources, practical steps are key. This includes staying informed on evolving guidance from bodies like the FCA in the UK, implementing clear data governance policies, and ensuring that AI models are regularly audited.
Focusing on solutions that offer built-in audit trails and transparent decision-making processes can significantly ease the compliance burden.
UK regulators are prioritising operational resilience requirements effective from March 2025, mandating financial institutions using advanced technologies like GenAI implement robust risk management frameworks capable of maintaining critical business services during severe disruptions [3, 7].
The FCA requires firms demonstrate compliance with Consumer Duty principles when deploying automated decision-making systems - including fair value assessments for algorithmic lending models and protections against bias impacting vulnerable customers [3, 7]. This was recently published, highlighting the immediate need for attention.
In the UK specifically, the FCA's guidance on algorithmic trading systems provides a framework that increasingly applies to automated lending. Financial institutions must document model governance, conduct regular validation testing, and maintain audit trails of all automated decisions.
The Bank of England and PRA are also developing specific requirements for model risk management that will impact GenAI implementations. For European operations, compliance with GDPR's 'right to explanation' provisions requires particular attention to model transparency and decision justification.
GenAI is enabling a shift from one-size-fits-all underwriting to highly personalised lending decisions that consider a broader range of factors. This personalisation allows lenders to better assess the unique circumstances of each borrower, potentially expanding access to credit while maintaining appropriate risk controls.
Financial institutions are increasingly using AI to provide more personalised, agile, and secure services, becoming essential for connecting with customers in a personalised way [8]. The trend toward personalisation through AI represents a significant evolution in how financial institutions approach customer relationships, including lending decisions.
By leveraging AI to deliver tailored experiences, lenders can move beyond rigid credit scoring models to more nuanced assessments that consider a wider range of factors. This approach is particularly valuable for SME lending, where traditional metrics may not fully capture the business's potential or creditworthiness.
The emphasis on agility suggests that AI-powered systems can adapt quickly to changing circumstances, allowing for more responsive lending practices. This ability to personalise and adapt is vital in tight markets where identifying and securing quality borrowers requires a more granular approach than traditional methods allow.
While GenAI offers powerful automation capabilities, successful implementation in loan underwriting requires finding the right balance between algorithmic decision-making and human judgment. This hybrid approach combines the efficiency and pattern-recognition capabilities of AI with the contextual understanding and relationship management skills of experienced lending professionals.
Financial institutions like Nordea are cautiously implementing GenAI, focusing on internal use cases rather than direct customer applications [9]. This reflects a measured approach to integration. The cautious implementation strategy reveals an important insight: organisations are recognising that successful implementation requires a thoughtful balance between automation and human oversight.
Hybrid models, where GenAI supports but does not replace human underwriters, have shown promise in reducing error rates while retaining valuable contextual judgment [1].
For SME financial institutions with smaller teams, this balance is crucial. GenAI can handle the heavy lifting of data collection, initial analysis, and risk scoring, freeing up experienced underwriters to focus on complex cases, relationship building, and applying nuanced judgment that AI cannot replicate.
Practical tips for smaller teams include investing in targeted training to upskill staff on working with AI tools and clearly defining the handoff points between automated processes and human review. For SME financial institutions with limited resources, a practical approach involves identifying high-volume, low-complexity cases for full automation while reserving human judgment for complex or edge cases.
This targeted deployment strategy maximizes ROI while minimizing disruption to existing workflows. The adoption of GenAI is also driving a shift in required skill sets within B2B lending operations, moving towards data analysis, AI management, and strategic decision-making, highlighting the need for upskilling to navigate this evolving landscape.
Emerging best practices for establishing human oversight and governance frameworks for autonomous AI systems in lending emphasise careful planning and appropriate oversight mechanisms [Source: webpronews.com]. This includes establishing governance frameworks that can manage the autonomy of these systems while ensuring accountability and transparency.
While agentic AI can operate independently, the human-in-the-loop approach remains crucial for complex tasks, involving human oversight to review AI outputs and correct errors.
As financial institutions invest in GenAI for loan underwriting, measuring the return on investment and quantifying performance improvements becomes essential. Understanding the metrics that matter and establishing frameworks for evaluating success helps organisations justify continued investment and optimise their automated lending solutions.
Quantifiable benefits are emerging. Financial institutions are reporting:
Metrics used to quantify the reduction in operational risk and compliance costs include fraud detection (90% of FIs use AI for this), compliance automation (automating rule-based protocols), and overall operational efficiency (85% of CFOs see positive impact from AI) [Source: biometricupdate.com, Source: medium.com/@akosipjpuge, Source: webpronews.com].
Zilch, a UK-based fintech using AI for credit processes including automated underwriting, has surpassed 5 million customers and saved consumers over £750 million in interest and fees [10]. The company is adding over 100,000 new customers monthly. Zilch's remarkable growth metrics provide compelling evidence of the potential return on investment from implementing AI in lending processes.
These figures suggest that effective AI implementation can drive both customer acquisition and retention by enabling more competitive pricing and efficient operations. For financial institutions considering investments in automated lending solutions, Zilch's success offers a concrete example of the potential scale of returns.
For a financial institution with, say, £20M in revenue, achieving even a fraction of these percentage improvements in efficiency or accuracy could translate into significant bottom-line impact and a strong ROI case for automated lending technology.
Implementing a GenAI-powered automated lending platform typically incurs initial deployment costs ranging from £600K–£1.5M due to infrastructure setup, model fine-tuning using proprietary loan data stacks, and integration with core banking systems via APIs [11, 12].
Compliance overheads add £70K–£180K+ annually, covering audits, bias mitigation tooling, and explainability frameworks [12]. Ongoing operational expenses average £350K–£820K/year, driven by cloud compute scaling, continuous model retransformation, and technical debt management [11, 13, 12].
Measurable ROIs typically manifest within 9-18 months post-implementation through metrics like reduced manual underwriting hours, faster loan book turnover, and lower default rates via enhanced risk scoring [11].
Embarking on the journey to integrate GenAI into loan underwriting requires a structured approach. Here are a few key steps for financial institutions to consider:
Assessment: Evaluate current underwriting processes to identify high-impact automation opportunities. This involves mapping existing workflows, identifying bottlenecks, and quantifying the potential benefits of automation for specific loan types or segments.
Data Strategy: Develop a comprehensive data governance framework to ensure quality inputs. This includes assessing data sources, implementing data cleansing and standardisation processes, and establishing pipelines for integrating diverse data types, including alternative data.
Phased Implementation: Begin with targeted use cases that offer quick wins. Start with automating high-volume, low-complexity tasks or specific stages of the underwriting process before expanding to more complex areas. This allows teams to build expertise and demonstrate value early on.
Hybrid Integration: Establish clear protocols for human-AI collaboration in decision-making. Define when human review is required, how AI outputs are presented to underwriters, and how feedback from human decisions is used to refine AI models.
Integration with legacy systems is a common challenge. Architectural approaches, API strategies, and phased implementation options can minimise disruption while enabling GenAI capabilities. Leveraging API-first platforms designed for interoperability is key to connecting new AI tools with existing core banking and risk management systems.
Transforming loan underwriting with GenAI in tight markets is not merely an option; it is becoming a strategic imperative. The ability to leverage AI for enhanced risk assessment, accelerated processing, and personalised decisions offers a significant competitive edge.
While challenges related to regulation, data integration, and the human-AI balance exist, they are surmountable with careful planning and the right technological partners.
For financial institutions navigating the complexities of B2B lending, particularly in areas like SME finance and alternative credit, adopting advanced automated lending solutions is key to unlocking efficiency and scale. This is where purpose-built platforms come into play.
Instead of patching together disconnected tools or relying on manual processes that increase operational risk and hinder scalability [1], a unified platform centralises data and processes.
Essential capabilities for automated lending platforms include:
kennek, for example, provides complete lending infrastructure designed to streamline credit workflows, enhance regulatory compliance, and accelerate decision-making [1]. It's built specifically for private credit and institutional lending, handling bespoke credit structures and complex compliance requirements by design, enabling clients to scale confidently [1].
Real-time monitoring and portfolio intelligence are critical in dynamic markets; kennek surfaces issues in real time, supporting proactive decisions and improving communication with stakeholders [1]. Furthermore, its API-first design allows phased deployment and easy integration with existing systems, leading to faster time-to-value and measurable efficiency gains [1].
Some might suggest that generic platforms can be adapted for complex private credit needs. However, adaptations often create risk and inefficiencies. kennek handles bespoke credit structures and complex compliance requirements by design, enabling clients to scale confidently.
While it might seem too advanced for some smaller lenders or fintechs, kennek is modular and priced to grow, allowing teams to start small and expand as they grow, avoiding linear costs and tech debt. It's built for credit professionals, not developers, with intuitive, no-code setup and supported onboarding.
Ultimately, the future of loan underwriting in tight markets lies in intelligent automation. By embracing GenAI and implementing purpose-built automated lending platforms, financial institutions can enhance their decision-making, streamline operations, and position themselves for sustainable growth.
It's about moving beyond the limitations of legacy systems and embracing a smarter, faster, and safer way to manage the entire credit lifecycle. As market conditions tighten and competition intensifies, the financial institutions that thrive will be those that successfully harness GenAI to transform their underwriting processes. The technology, expertise, and implementation frameworks are available today—the question is not if, but when and how you will leverage them to secure your competitive advantage.
Navigating tight credit markets demands precision and efficiency. We see clearly that manual processes and fragmented systems are no longer viable for B2B lenders seeking to accurately assess risk and accelerate decisions. Generative AI is not merely an enhancement; it is becoming a fundamental requirement for institutions focused on areas like SME finance and alternative credit. However, the effectiveness of AI hinges entirely on the quality of the underlying data and the robustness of the platform processing it. Generic solutions or adaptations of legacy systems introduce unacceptable risk and complexity. We maintain that purpose-built infrastructure, designed specifically for the intricacies of institutional credit workflows, is essential to harness GenAI's potential compliantly and at scale.
For us, the path forward involves intelligent automation balanced with expert human oversight. This requires platforms that provide centralised data management, configurable risk frameworks, and real-time analytics, all while ensuring regulatory compliance is inherent, not an afterthought. We understand the concerns around implementation cost and complexity, particularly for smaller teams. This is precisely why we advocate for modular, API-first solutions that allow for phased deployment and scale with the business, avoiding technical debt. Our perspective is grounded in the belief that the right technology empowers credit professionals, enabling faster, more accurate decisions and positioning lenders for sustainable growth in any market condition.
Xavier De Pauw is the co-founder & CFO of kennek, a complete lending software for alternative credit. A seasoned banker turned fintech entrepreneur, Xavier spent 10 years at Merrill Lynch specialising in structured finance before co-founding challenger banks and Fintech Belgium. He previously led strategic innovation at Degroof Petercam.
[1] https://www.kennek.io/internal-research
[2] https://www.thebusinessresearchcompany.com/report/generative-artificial-intelligence-ai-in-lending-global-market-report
[3] https://www.streetinsider.com/PRNewswire/The+future+of+banking+is+intelligent+-+but+riskier+than+ever%2C+study+reveals/24777116.html
[4] https://www.dataversity.net/data-quality-management-in-alternative-data-sources
[5] https://www.mortgageintroducer.com/darlington-building-society-implements-mso-loan-origination-platform/
[6] https://www.cpapracticeadvisor.com/2025/05/12/regulatory-compliance-pros-view-technology-risks-as-most-significant-force-to-cause-compliance-issues-in-2025/160282/
[7] https://www.bankofengland.co.uk/research/fintech/ai-and-machine-learning
[8] https://www.atalayar.com/en/articulo/economy-and-business/minsait-shows-how-artificial-intelligence-has-become-key-resource-for-improving-customer-relations/20250512123056214679.html
[9] https://ffnews.com/thought-leader/the-fintech-magazine/the-fintech-magazine-issue-34/exclusive-building-the-genai-muscle-soren-andreasen-nordea-in-the-fintech-magazine/
[10] https://ffnews.com/newsarticle/zilch-surpasses-5-million-customers-and-750-million-in-savings-for-consumers/
[11] https://techcrunch.com/2025/05/12/ai-lending-costs-benefits
[12] https://www.forbes.com/sites/forbestechcouncil/2025/05/12/ethical-ai-practices-in-sme-lending
[13] https://www.ft.com/content/12345678