Insights at a Glance
In a challenging credit environment, better lending technology is critical to bank CIOs’ impact:
Meanwhile, they feel pressure to apply AI to improve business outcomes:
Banking and investing IT leaders cite AI implementation and systems modernization among their top three priorities (51% each). Thirty-six percent of respondents lag expectations for AI implementation; likely contributing factors include accelerated AI evolution and hype, a daunting range of use cases, and the opacity of embedded AI features.
Gartner identifies three opportunities for bank CIOs to succeed with AI in lending. The most immediate benefit comes from:
This research note addresses the medium-term opportunities to guide bank CIOs in piloting AI for:
A future-looking strategy requires assessing AI agent use cases for:
Issue Context
The 2025 Gartner Business Outcomes of Technology Survey underscores how important lending technologies are to bank CIOs in an uncertain, competitive lending environment:
Fifty-nine percent of senior commercial banking IT decision makers say loan management technology is extremely important for meeting bank goals. Forty-nine percent in retail banking say that about origination technology.
In retail banking, IT decision makers cite credit decisioning analytics and loan origination systems as among the top three technologies for deployment within the next two years. In commercial banking, loan origination and trade and supply chain finance solutions are among the four most frequently cited technologies.
Meanwhile, bank CIOs feel growing pressure to apply technology — namely AI — to improve business outcomes, as per the 2H25 Gartner Financial Services Leaders’ AI and Priorities Survey:
The implementation of AI capabilities is tied to systems modernization as the top-three priority most cited by banking and investment services IT leaders.
Thirty-six percent, though, lag behind expectations for progress in implementing AI.
The lag is concerning, since AI can enhance loan origination and management, and vendors are incorporating AI features. Multiple factors, however, make AI prioritization and implementation challenging for bank CIOs: the accelerated evolution and hype, the complexity of advanced AI capabilities, the lack of clear objectives, the opacity of embedded AI features, and the daunting range of use cases. CIOs, therefore, risk making suboptimal decisions that can result in:
Failure to deliver promised benefits, such as a reduction in delinquencies, faster time to market, or access to new borrower segments.
Unnecessary use of advanced (and expensive) AI when simpler, cheaper approaches would suffice.
After applying AI to transform document-dependent processes, bank CIOs can drive value in lending by piloting predictive AI and synthetic data to enhance credit risk assessment and support debt recovery.
Impact Brief
Bank CIOs can maximize impact by weighing AI use cases in terms of business outcomes that range from improving credit process efficiencies to creating innovative financing solutions tailored to complex client needs. After applying AI to document-dependent processes, CIOs should turn to credit risk assessment and management as their next use case for AI adoption.
So far, lenders are not using AI models alone for final credit decisioning, given regulatory, governance and risk considerations. However, loan origination and loan management solutions are starting to incorporate AI to support credit decisioning and risk management. These advances point to three areas of credit risk where CIOs should prepare to pilot predictive AI, synthetic data and other advanced AI techniques (see Table 1):
AI Techniques and Approaches | Use in Loan Origination and Loan Management |
Probabilistic AI-Based Models | To complement and support deterministic rule-based credit decisioning |
Synthetic Data | To fill data gaps and model risk in new lending markets or for new segments |
Predictive and Generative AI | To improve early warning default prediction and debt recovery |
|
Source: Gartner (November 2025)
More Detail
Support Deterministic Rule-Based Credit Decisioning With Probabilistic AI-Based Models
Modern credit decision solutions have begun to complement deterministic rule-based decisioning with probabilistic AI-based models. One approach to doing that is through PMML (predictive model markup language), which effectively helps “translate” complex AI scoring models and introduce them into rule-based decisioning systems.
As commercially available loan origination and management solutions increasingly incorporate AI-based risk management features, bank CIOs can work with business leaders, chief data and analytics officers and data scientists to assess AI techniques to complement rule-based credit decisioning. Such use cases for AI include:
Vetting credit decisions.
Enhancing the credit decision engine input through data deduplication and improved data extraction.
Processing larger risk datasets, including alternative data, to improve decisioning accuracy and expand credit to underserved customers.
Refining credit attributes through improved discovery of nonlinear data relationships, thereby improving decisioning precision.
It’s important to note, however, that the return on investment from the use of AI to enhance existing credit decisioning models is still an open question. For instance, any resulting shift in the Kolmogorov-Smirnov (K-S) statistic, used to validate a scoring model’s ability to separate bad from good risks, may end up being too negligible to justify the potentially significant cost of AI models.
As AI begins to support credit decisioning, bank CIOs will need to align data scientists to enable the use of explainable AI (XAI). That will help ensure their banks are in compliance with regulations by explaining credit decisions and identifying and mitigating model bias.
Representative vendor offerings deploying AI to support credit decisioning include the following:
Centelon’s Finnate loan origination and management solution enhances traditional rule-based assessment of borrower eligibility with ML-driven analysis and mapping to bank credit policies (used in model prompting). The output is a decisioning recommendation for the underwriter.
Newgen’s credit decision engine combines a probabilistic AI model battery for assessing alternative underwriting data with a traditional deterministic rule engine.
Nucleus Software’s loan origination solution enables AI-based confidence scores for loan applications to support underwriters in credit decisioning. The model is trained on anonymized historical data trends that include customer behavior, demographics, financial information, and geographical loan trends, and can be configured to align with bank credit policies and customer data.
Use Synthetic Data to Fill Data Gaps and Model Risk in New Lending Markets or Segments
Lack of sufficient or high-quality data has an impact on revenue growth, hindering the extension of loans to small businesses or underserved customer segments. Missing values, outliers, or underrepresentation of certain segments, such as recent immigrants, make it challenging to build effective credit decisioning models or adapt loan products for these segments.
While the use of synthetic data in lending is not yet widespread, bank CIOs can direct data scientists to test synthetic data’s potential to accelerate loan growth, expand access to new customers and increase financial inclusion through its ability to enrich data for underwriting. Synthetic data can mimic various types of data such as customer profiles, transactions or user behavior and can be used in a variety of ways to improve risk assessment and expand revenue opportunities, such as:
Plugging gaps in existing datasets, for instance, to enhance the sample of underrepresented customer segments, such as those who are financially vulnerable.
Simulating new market conditions or behavioral changes in response to unexpected events, thereby facilitating the adaptation of credit products.
Aiding the development and testing of new credit products and risk models in a controlled environment without exposing customer information. These credit risk models can be based on alternative underwriting data, such as cash flow and on-time payment of rent and utility bills, thus expanding credit to the underserved.
Bank CIOs must understand the accuracy, transparency and compliance risks, which have constrained the use of synthetic data in lending to date:
Synthetic data can miss natural anomalies, add complexity to model development or even fail to contribute any materially new information. Paradoxically, while it can help address biases affecting underrepresented groups, there is a risk that the underlying data used to generate it could also contain inherent biases.
The models that generate synthetic data lack transparency, which means any subsequent AI models trained on that data, in whole or in part, will lack explainability.
Synthetic datasets may appear realistic and accurate, regardless of whether they accurately capture the underlying real-world environment. The lack of expertise in synthetic data generation and the difficulty of advanced synthetic generation techniques like generative adversarial networks (GANs) can impede progress.
Improve Early Warning Default Prediction and Debt Recovery Through Predictive and GenAI
Predictive and generative AI (GenAI) play a growing role in better collections processes, given their ability to enhance the sensitivity of default forecasting and to support debt recovery strategies. Advances that impact how banks increase the odds of repayment include:
AI-enhanced predictive analytics supports early warning systems (EWSs) that can flag the likelihood of loan delinquency several weeks or even months out. Such data can then guide the type and timing of outreach and intervention.
GenAI can support sentiment analysis of interactions with borrowers, which can help gauge customer likelihood to repay as the loan nears or enters delinquency. It can also be used to propose conversational approaches and next best actions with borrowers, help generate call scripts, suggest suitable outreach cadence and channel, and support collections staff assignments based on profile and workload.
For high-volume collections, CIOs can direct IT to vet GenAI-powered voice bots that can make calls to overdue borrowers, saving staff from spending inordinate amounts of time trying to contact them. Such bots can even negotiate payment options and customize the conversational tone.
AI-based debt recovery offerings, either as stand-alone software or as credit management modules, include solutions from vendors such as Aurionpro, JurisTech, Loxon, Newgen, Nucleus Software, Pennant Technologies, Qualco, and Skyworx Indonesia.
Recommendations for Bank CIOs
Supervise an initiative to assess the shortcomings of the bank’s current credit decisioning software (such as limited rules configurability or overly narrow credit attribute sets). Coordinate with data scientists on the requirements for developing a probabilistic AI model to complement the credit decision engine, identifying architectural requirements as well as potential vendors.
Direct IT to confer with credit underwriters about data-poor, untapped credit opportunities or gaps in other risk datasets. Provide these findings to data scientists to determine the feasibility of generating synthetic data to fill those gaps while putting guardrails in place to protect against potential biases.
Lead a business-IT review of the bank’s current loan management solutions for features like early default warnings and collections process orchestration, looking for opportunities to introduce predictive and generative AI for gauging likely delinquencies and supporting debt recovery.
2H25 Gartner Financial Services Leaders’ Priorities and AI Survey. This survey sought to understand financial services leaders’ most pressing priorities for the year, as well as AI initiatives at their organizations. This survey was conducted online from 30 September through 3 November 2025. In total, 158 executives at financial services organizations participated, including 82 banking and investment services respondents and 56 insurance respondents. Respondents were located in North America (n = 71), Europe (n = 35), Asia/Pacific (n = 16), the Middle East and Africa (n = 11), and Latin America and the Caribbean (n = 5). Disclaimer: The results of this survey do not represent global findings or the market as a whole, but reflect the sentiments of the respondents and companies surveyed.
2025 Gartner Business Outcomes of Technology Survey. This survey was conducted to understand how industries leverage technologies for various use cases. It assessed investment, deployment and implementation strategies for industry technologies. It also examined key areas intended to be impacted by technology investments, including challenges to realizing business outcomes and industry key performance indicators. The survey was conducted online from June through August 2025. The 648 respondents were from midsize, large and global enterprises from North America, EMEA and Asia/Pacific. The respondents were screened for senior IT and some business leadership roles with technology decision-making responsibilities. Disclaimer: The results of this survey do not represent global findings or the market as a whole, but reflect the sentiments of the respondents and companies surveyed.