blog

5 Ways AI Impacts Structured Finance

dylan.thiam@algoritmica.ai (Dylan Thiam) — Mon, 03 Aug 2026 12:48:01 GMT

It feels like the 1980s: analysts are skeptical—Lotus 1-2-3 is their trustworthy spreadsheet, but Excel will soon transform how they get work done. Fast-forward to today, and the structured finance industry is facing a similar shift with the recent advancements in AI.

New AI models are unlocking patterns hidden in vast datasets, automating processes, and revealing insights that often remain in our blind spots. But that’s not the only problem you can solve with AI.

Let’s take a look at 5 key areas of structured finance where AI can bring tangible benefits.

Repetitive and Manual Processes: Many workflows, such as cash flow modeling or asset-level analysis, are manual, repetitive, and prone to errors. A lot of the work involves plugging in assumptions, running calculations and tweaking things over and over for different scenarios.

Imagine taking this whole process and automating it. Take stress testing as an example: an AI model can simulate thousands of scenarios in minutes, a great deal faster than traditional models. Advanced pattern recognition could also help catch potential issues early.
Incomplete and low-quality data: The sector depends on large volumes of data that are often incomplete or riddled with issues, which undermines the reliability of forecasts and increases risk exposure. Important insights can be buried under inconsistent or incomplete data.

Analysts are using AI to clean and prepare unstructured data for analysis. Automated cleaning—deduplication, flagging/removing missing fields, formatting, etc.—makes life a lot easier. Especially in a sector where data may come from multiple sources, AI can simplify the data integration process.
Measuring Unknowns in Portfolio Risk: Portfolio risk analysis can depend on identifying complex patterns in diverse datasets. Traditional models rely heavily on historical data and linear assumptions, failing to capture the more nuanced and dynamic risks of today’s markets.

Stress tests also require modeling multiple inputs and more granular analysis at the asset level, and today’s tools are hard to scale for complex scenarios.

AI models are able to identify correlations between asset performance and macroeconomic indicators, and can integrate non-traditional data sources, offering more reliable predictions.

Generative AI can further provide clear visualizations of stress test results and generate scenario-specific reports for different audiences.
Compliance & Reporting: Legal teams are buried under heaps of documentation with exhaustive details about the deal: 200+ pages of prospectus to read in order to identify key terms and potential issues. And that’s before monitoring for regulatory updates to ensure continuous compliance.

Teams can analyze prospectuses through a smarter interface. Imagine having an assistant that can read the prospectus in minutes, and quickly answer every question while referencing the original document. One example of this is Prospekto, a tool helping professionals quickly make sense of Asset-Backed Securities (ABS) deals.

Taking it a step further, users could quickly generate compliance reports using data from multiple sources. Secondly, AI can scan and interpret regulatory changes in real-time, helping compliance team stay up to date.
Emerging Risks: Trends such as ESG-focused investments and electric vehicle (EV) financing are introducing new datasets and risk factors that traditional models aren’t equipped to handle.

AI, on the other hand, can handle unconventional data sources, such as ESG ratings, climate impact assessments or EV charging infrastructure data. Thus, it enables a more realistic and accurate prediction of asset performance and borrower risk. It is great for analyzing new relationships—e.g. residual value trends—and uncovering patterns that would escape traditional models.

Barriers to Adoption

The structured finance sector operates at the intersection of complexity and opportunity. While the benefits of AI are compelling, the path to adoption and success is challenging:

Interpretability and Trust: AI models are often seen as “black boxes,” making it difficult for investors to trust their outputs. Decision-making is high stakes in structured finance, meaning stakeholders need to be able to trust and justify the rationale behind predictions.

This lack of transparency becomes an even greater challenge when it comes to sensitive user data, which is another reason many financial institutions hesitate to adopt AI. The sector needs transparent models to build confidence in AI outputs.
Better Access to Clean Data: AI models rely on high-quality data to generate accurate insights, but structured finance datasets are often fragmented, inconsistent, or incomplete. Without clean, well-structured data, even the most advanced AI models can produce unreliable results.
Cost Perception: Some institutions see AI as a luxury, a tool for firms with big budgets instead of a necessity. Smaller players sometimes feel they lack the scale to justify the investment. This perception is slowly changing with the rise of affordable and accessible tools. In this sense, the barrier isn’t the actual cost as much as the perceived ROI.
Regulatory Compliance: The structured finance sector is highly regulated, and AI tools have to meet strict compliance standards. On one hand, organisations have to prove that AI recommendations align with regulations. On the other, emerging regulations about AI itself add another layer of complexity. No one wants a technology that could cause compliance headaches in the future.
Complexity of Implementing: Integrating AI into existing workflows might require significant time, resources, and technical expertise, particularly for firms unfamiliar with the technology.
In-house Expertise: AI adoption requires skilled professionals who can train and fine-tune models, interpret AI outputs and bridge the gap between technology and business needs. Lack of expertise might lead to poor use of the technology or hesitation.

Truth is, structured finance has always been a high-stakes industry, and change won’t come easily—just like for our analysts back in the 90s. Today, data issues, infrastructure, cost concerns and regulatory hurdles are all real challenges.

For most organizations, adopting AI will not happen overnight. But Excel wasn’t an immediate replacement; it worked alongside existing tools and gradually proved its value in certain areas before it became a necessity.

The potential is impressive, and we may be looking at serious changes in both the tech stack and the company culture of institutions in this space—sooner than we might think.

For more on this topic, subscribe to this blog!

Structured Finance Hackathon 2026: Recap and Key Highlights

luca.borella@algoritmica.ai (Luca Borella) — Mon, 03 Aug 2026 12:45:13 GMT

Hi there!

The third edition of the Structured Finance Hackathon brought together more than 80 developers, data scientists, AI builders and finance professionals to tackle some of the industry’s most difficult technology challenges.

Following two weeks of online collaboration, seven finalist teams presented their prototypes on June 10 at Torre Diagonal in Barcelona. Their projects explored how open-source technology, data and AI could improve credit modelling, agentic workflows and the financing of a new generation of physical assets.

Catch the highlights in our event recap video.

Hackathon Winners & Highlights

Innovation was at the forefront of this year’s projects. Here below are listed the best solutions for each hackathon challenge:

World Model for Credit: The Winning team created a framework for training a foundational credit model. The framework includes ingestion, tokenization, training, and downstream evaluation pipelines. After their initial iteration, the team is upgrading its training capabilities leveraging the NVIDIA Incubation program, using eight NVIDIA H100 GPUs, providing 640GB of combined VRAM for a 60-day development period.

Agentic Framework for Structured Finance: This challenge was solved by LoanWhiz and the Hypoport teams. Both teams presented approaches combining deterministic financial components with LLM-powered orchestration. Rather than asking a language model to perform every task, their approaches use specialised and testable components for activities such as extraction, validation, calculation and evidence generation. This hybrid architecture can help improve reliability, traceability and human oversight when AI is used in capital markets workflows. Going forward we will nurture an open-source ecosystem of primitives and explore the opportunity to use Fluxnova, a leading FINOS project, as orchestration engine.

Data Infrastructure Financing: Greyhound responded by addressing the static nature of collateral monitoring. It provided an investor-facing interface for datacenter finance that replaces periodic audits with continuous, source-verified data feeds (demo video). We are seeking trial users to help test workflows and shape the future of these prototypes. If you work for an asset manager, an asset owner or a lender consider joining the Pilot Program.

Expert Panel

Following the project demonstrations, we asked a broader question: What would it take for these prototypes to become trusted, production-grade infrastructure across banks, originators, investors and rating agencies? The discussion was moderated by Edward James (RCQ Associates), featuring:

Daniel Goudsmit, Business Development Lead, Hypoport
Frank Benhamou, Head of SRT at Cheyne Capital and Co-founder of Tranchify
Hanna Sundqvist, Head of Private Credit, Moody’s Analytics
Manolis Manoli, Treasury, N26

The panel explored the importance of data quality, deterministic financial logic, human oversight, explainability and governance when applying AI to structured-finance workflows.

Looking Ahead

Before turning our attention to the next edition, it is worth recognising the cumulative value created by an openly collaborative environment. The 2026 event was the third Structured Finance Hackathon, and we are beginning to see participants build on projects, data models and lessons from previous editions.

For example, CreditGoose, originally developed during the 2025 hackathon, returned this year as part of the Goose training session for quants, risk professionals and front-office practitioners in partnership with OpenRouter and Barcelona Code School. CreditGoose project provided a practical example of how an open source prototype can be reused, extended and connected to new governance and agent-development patterns.

This is the model we want the hackathon to support: Prototype → Open-source project → Community validation → Industry pilot → Production-ready infrastructure

Thank You!

A huge thank you to our sponsors for their support:

Channel Capital sponsored the World Model for Credit challenge, asking teams to explore how payment behaviour – including arrears, cures, defaults, recoveries and prepayments – could be transformed into reusable credit representations. Kris Wilson, Channel Capital’s Chief Operating and Technology Officer, joined us in Barcelona and served on the judging panel. We sat down with him and discussed the practical requirements for adopting AI in private credit, including strong data foundations, governance, human oversight and organisational adoption
Hypoport: Hypoport sponsored the Structured Finance-Native Agent Framework challenge and brought extensive expertise in structured-finance data, reporting and industry workflows. The challenge asked participants to move beyond standalone AI applications and build reusable components (i.e. primitives) for working with prospectuses, ESMA loan-level data, remittance files, trustee reports, waterfall logic and other structured-finance information. Daniel Goudsmit, Business Development Lead at Hypoport, also joined our expert panel to discuss what is required to move AI prototypes into trusted industry infrastructure
FINOS: Fintech Open Source Foundation (FINOS) provided essential best practices for collaboration and AI governance in regulated enterprises. Hackathon participants used the AI Governance Framework (AIGF) as the governance backbone of their projects, bridging the gap between their prototypes and potential future production-level applications.

We are particularly grateful to AI Tinkerers, viaNexus, RCQ Associates, zetta10, n8n, BCN FinTech, and Scaleway, the judges, speakers, instructors and volunteers who contributed their time and expertise for supporting the event and its community.

Most importantly, thank you to every team that was willing to build openly, share its work and demonstrate what the next generation of structured-finance technology could look like.

Stay tuned – there is much more to come!

AI Meets Structured Finance: Insights from the Hackathon

dylan.thiam@algoritmica.ai (Dylan Thiam) — Mon, 03 Aug 2026 12:40:46 GMT

For an industry that still uses fax machines in places, “hackathon” might not be the word that gets most structured finance professionals out of bed in the morning. But the Structured Finance Hackathon 2025, which wrapped up in Barcelona on June 12, suggests that might soon change.

Held in the appropriate setting of the former Barcelona Stock Exchange, this hybrid gathering brought together developers and finance insiders to do something our industry rarely dares: build things quickly.

The result? Prototypes that make a strong case for a new generation of tools – and talent – in structured finance.

Below, I walk through the standout projects. If you’re wondering whether structured finance could ever get a little more modern, a little more intelligent (and yes, a little faster), the following might convince you it’s on the right track.

Climate Risk Insights of Dutch Mortgage Portfolios

🏆 First Place Winner

View Project

It’s hard to make the phrase “Dutch mortgage climate exposure” sound exciting, unless you’ve just built a multi-agent AI system that reads loan books, cleans zipcode data, assesses regional flood risk, and turns all that into a dashboard that makes sense.

This team did just that, targeting the €800bn blind spot in Dutch mortgage risk: climate events like the 2021 Limburg floods that banks didn’t price for.

This smart orchestration of agents (data reader, zipcode processor, climate calculator, dashboard generator) can help banks identify key vulnerabilities in mortgage portfolios before climate events occur.

AI-DRRD – AI-Powered Disaster Risk Resilience Dashboard

🥈 Second Place (Tied)

View Project

This is another dashboard you wish you had in 2021: a risk dashboard with multiple hazard types, property-level detail, scenario modelling, and automated regulatory reporting. AI-DRRD connects climate data, even real-time weather information, with loan portfolios in a way that’s granular, visual, and financially legible.

The AI-DRRD visualizes climate risks at the property level to enable more accurate loan pricing and effective risk mitigation. The AI Assistant provides real-time guidance on mitigation strategies and overall portfolio optimization.

It wasn’t just the tech that stood out. The presenter behind AI-DRRD delivered one of the most memorable demos of the day. Anchored in a sharp understanding of both climate risk and financial impact, her storytelling made a complex tool approachable. A reminder that how you explain the solution can be just as important as the solution itself.

Credit Goose – Smarter Invoice Financing with Agents

🥈 Second Place (Tied)

View Project

For anyone on the lending or credit side who’s seen promising SMEs fall through the cracks due to inadequate scoring engines, Credit Goose offers a pragmatic fix.

The platform uses a Goose-powered AI agent to score invoices based on both financial and public data in just 45 seconds. SMEs connect CreditGoose to their ERP to retrieve unpaid invoices, the agent then evaluates the risk score and enables financing to be disbursed in 24 hours.

SMEs can set up auto-finance for eligible invoices based on certain criteria
3-stage scoring framework: Qualification Filter, Risk Bucketing, Intelligent Scoring
Real-time Web Search using Gemini LLM to find updated public information
Admin panel for tweaking queries without touching code

InvoiceAccelerator – Headless AI for Structured Workflows

View Project

InvoiceAccelerator exposes a Goose-based agent service as a HTTP API. Its LLM backend is Groq-powered for low-latency inference (yes, that’s the same Groq from the AI speed demo videos), and it turns invoices, borrower data, and risk parameters into financing decisions.

The team focused on integration over interface. By exposing everything via a clean API, the system is UI-agnostic, ready to plug into an existing dashboard or support a custom-built interface. This is a B2B SaaS backend in the making: composable, pluggable, and with no frontend dependencies.

Issuria – Natural Language to PRIIPs KID

View Project

Ever wanted to type “3-year note on tech stocks with 8% max return” and get a compliant Key Information Document (KID) back? This submission makes that possible. Issuria’s AI-powered assistant takes plain language product specs and spits out export-ready documents.

With integration-ready CLI tools, HuggingFace-hosted datasets tailored to ESMA/CDM benchmarks, and a multi-agent backend coordinating compliance, structuring, and product creation, Issuria cuts down the issuance cycle from weeks to minutes.

Converts freeform product descriptions to structured JSON
Supports document editing and .docx export
Uses Chronos model for synthetic risk data
FINOS Legend for compliance logic; integrates OBP APIs

Somewhere, a structuring desk just sighed in relief.

TWBRIYPR – The Midnight PSP Fixer

View Project

Built by the TWBRIYPR (This Will be Reflected in Your Promotion Rationale) team, this one solves a deeply relatable problem: it’s 2am, your PSP is down, and your only recourse is a sleepy engineer with database access. This tool replaces panic with logic by letting payment specialists (not developers) define fallback rules in a shared Google Doc. AI reads the document and acts accordingly.

The concept of a human-readable config layer for routing logic (editable by specialists and not just engineers) is broadly applicable even outside the original hackathon scope.

It makes you wonder what other frustrating manual SOPs are waiting to be replaced by plain-text + LLMs.

Reflections from the Floor

Yes, there were some technical hiccups – mainly the kind that plague hybrid events from streaming issues to audio glitches. What stood out was the patience and good humour with which attendees took it all in stride.

One thing the event reaffirmed: writing challenges for a finance hackathon requires more than just domain depth. The more complex the topic, the more important it is to translate the problem into technical hooks that developers can actually grab onto. The projects that went furthest were those that understood both sides of that bridge.

There’s something charming about watching devs who met six days ago build AI prototypes for such a complex sector.

Three takeaways for the structured finance professional:

Structured finance is attracting developers. Not hordes of them, but more than before. Many are still getting to know the finance sector, while others are ex-finance professionals themselves.
Agentic AI is coming to the back office. What began as a buzzword is quietly morphing into tools that reduce friction in underwriting, reporting, and product structuring.
These tools need finance experts. Innovators face a common challenge, i.e. to find relevant expertise to help with product design. If you’ve ever complained about software made by people who don’t understand your business… well, this is how you get involved.

What’s Next?

There’s talk of another edition, possibly later this year, or early next. If you’re reading this and thinking that these kinds of tools are light-years away from being built and adopted, consider this: these hackathon teams are already working on the first part.

With the right partners, good data, and a shared vision, structured finance doesn’t have to move at a glacial pace.