Fintech is rewiring the foundations of securitisation
Jacco Samuels, Managing Director, Hypoport
Daniel Goudsmit, Business Development Lead, Hypoport
Hypoport Executives on the AI Revolution in Securitisation
Digital infrastructure is quietly transforming the securitisation ecosystem. Hypoport, an Amsterdam-based fintech services provider in Amsterdam, is at the heart of that change. Managing Director Jacco Samuels and Business Development Lead Daniel Goudsmit discuss how tech made its mark and how artificial intelligence is set to reshape it once again.
Over the past decade technology, from blockchain to distributed ledgers, promised to reshape securitisation. Which initiatives actually delivered?
Daniel Goudsmit: The biggest shift wasn’t blockchain at all, it was the move to the cloud. Ten years ago, no Dutch bank wanted core banking or customer data anywhere near the cloud. Now, effectively all of them do. That change came very fast and it’s become the base layer for everything else we’re discussing.
Once data sits securely with a hyperscaler, it’s far easier to bolt on new tools and services. If your infrastructure is on Azure or similar, it’s straightforward for that provider to add features on top, whether that’s analytics, workflow or AI. The cloud has become the foundation for innovation in securitisation.
How has that changed day‑to‑day operations for issuers and servicers?
Jacco Samuels: It’s accelerated the whole software lifecycle. Cloud-based setups let you deploy updates faster and adopt new technologies more easily. For securitisation chains, especially in mortgages, this matters a lot. From origination to servicing, it used to be difficult to keep data consistent and in one place; now it’s much easier to track, share and reconcile the same data through the life of a deal. Blockchain, by contrast, barely registered in day‑to‑day securitisation. It was piloted in places but never really scaled. The quiet story has been cloud adoption and the operational flexibility that came with it.
Does that expansion of data access run up against privacy concerns?
Goudsmit: That’s where GDPR has played a surprisingly positive role. Around the same time that banks began thinking about the cloud, GDPR gave citizens a stronger sense that their data is governed and protected. That helped banks justify moving sensitive data into external infrastructure without clients feeling their information was just being handed over. You end up with two forces that look contradictory, centralised data and stricter privacy rules, but both are necessary. Without regulation, confidence in cloud-based banking infrastructure would have been much lower.
Where is AI already touching the securitisation process today?
Samuels: One of the clearest use cases is on the legal and documentation side: the prospectus, the offering circular, the waterfall logic. AI can help translate what’s written in those documents into the parameters the software actually needs.
Take the waterfall: it’s simply the ordering of payments to different stakeholders in a deal. Traditionally, that logic is written out in legal language and then re‑implemented manually in IT systems. AI can already assist by reading the legal text, extracting the right numbers and rules, and feeding them into performance and reporting platforms. That is happening now.
Goudsmit: A lot of securitisation is legal technique. AI is good at reading and summarising, and then turning those summaries into something operational. Instead of an analyst combing through a prospectus line by line, AI can pull out the quantitative parts, map them to the right fields and reduce errors and turnaround times.
How far do you push AI into the underlying data, given how sensitive it is?
Goudsmit: In many securitisations you have very granular loan-level data. Technically, that’s a rich field for AI. In practice, you run into privacy and sensitivity quickly. For a residential mortgage book, you’re not going to expose everything to an external model. There will be use cases where AI can work on anonymised or mapped datasets, and others where it simply isn’t suitable under current rules.
Samuels: One safe and powerful use case we already deploy is data mapping. If you need to translate one data model into another, AI can match fields and structures quickly. That’s not about exposing personal data broadly; it’s about aligning data points correctly. An experienced analyst can do it “by head,” but AI reduces the manual effort and frees capacity.
Is AI already influencing how deals are structured, not just how they are documented?
Goudsmit: Internally, we use AI tools as a kind of sparring partner. You can explore prompts around pool composition or triggers just to get your creative brain working. It’s not replacing the structurer, but it can surface angles you might not have considered.
Samuels: For example, optimising a pool against a set of constraints: say, certain triggers or target ratings. That’s something AI can help with. It can search the space of combinations quickly and flag good candidates. You still need human oversight and market judgement, but it changes how you approach the problem.
What exactly is Hypoport building in its AI “lab”?
Samuels: We think in layers. First, we focused on knowledge building: internal innovation days, expert workshops, and a central AI resource hub. Second, we moved into internal process automation, where we’re seeing the biggest gains right now.
We have around 55 employees, roughly half are software developers. They use AI continuously to write and review code, handle documentation and speed up development. We also rolled out a private AI environment — essentially our own instance of a large language model — integrated with our Jira helpdesk and Azure DevOps.
Why run your own model rather than using public tools?
Goudsmit: Privacy and control. Technically, it’s an instance of something like ChatGPT that we host locally, not connected to the public internet. That means staff can use it without worrying about client or proprietary data leaving our environment.
Samuels: We linked it to the helpdesk because client questions are a rich source of information. The system can learn from recurring queries and responses, helping us respond faster and making support more consistent. It also shows us where products or documentation need improvement.
How quickly has the AI curve steepened for you?
Goudsmit: In 2024 we ran an AI-focused innovation day and ran into all kinds of technical constraints: token limits, infrastructure issues, basic access. The environment wasn’t ready. A year later, the same event produced four, five, six ideas that teams could actually build on the day. That’s a huge difference in 12 months.
A real turning point for us was a hackathon at Global ABS, where someone demoed analysing a prospectus with AI in our own niche. Seeing it work on our kind of document made it clear this is no longer abstract.
Samuels: We also won a hackathon in Barcelona last year with a prototype that analysed climate risk in a mortgage portfolio using AI techniques. That was still “first layer” work — building experience and testing boundaries — but it showed how quickly these tools can be applied to real questions like climate exposure.
Will AI eventually write all the securitisation reporting?
Goudsmit: Technically, it’s hard to imagine that not being possible within five to ten years. You can picture a future where instead of static reports, a bank offers a chatbot: investors ask what they want, and the system produces a chart, table or narrative on demand. What slows that down isn’t the technology, it’s the regulatory framework and the comfort level of supervisors and investors. This is a heavily regulated market. The tech will outrun the rules for a while.
Samuels: There’s also a security dimension. AI makes it easier to detect vulnerabilities and also, unfortunately, to attack systems. Regulators are already requiring firms to prove how they manage security across the entire value chain, including suppliers. Companies that can innovate and still satisfy these demands will be the ones that gain share.
What does this mean for costs — and for the traditional fee pool in securitisation?
Goudsmit: If AI makes parts of the process more efficient, competitive forces among lawyers, advisors and other intermediaries like ourselves is inevitable and the total price of “a securitisation” could come down. Today’s securitisation market is relatively underdeveloped partly because each link in the chain is expensive. As AI drives efficiency, after the initial investment period, prices will come down.
In the long run, you can imagine a much more frictionless process, even something close to “push‑button” securitisation for standardised assets. That might raise uncomfortable questions about who captures value, but also opens the door to smaller, more frequent deals.
Samuels: You might end up doing more securitisations for the same total cost. It could make smaller portfolios economical and make it easier for new issuers to enter the market, provided you keep quality and risk management high.
Goudsmit: Covering for different risks or liabilities in the process is key. You hire lawyers and advisors not only for expertise but also for the insurance that comes with hiring their expertise. If you replace parts of that chain with AI, you still need to know who is accountable when something goes wrong. At some point you might see “AI plus liability” products emerge from big firms, but the liability question has to be solved.
Looking five years ahead, what are the most realistic AI-driven changes in securitisation?
Samuels: Every stage of the chain will be augmented. Data mapping, validation, legal translation, pool optimisation — AI will be embedded in all of it. That will help extend securitisation beyond highly standardised mortgage pools into less mature asset classes like SME and corporate loans, where today the manual effort is a barrier.
Goudsmit: At the same time, this is a conservative industry and that won’t change overnight as banks are built on trust. The immediate impact is efficiency and automation. The more interesting, longer-term impact is creativity: using AI as a tool for inspiration and problem-solving in a field that is usually very rigid.
Whatever the future holds – what’s cool right now is that technology is no longer a side panel at securitisation conferences. It’s in the main discussion now — and that, for us, is a very welcome development.
Register now
At the Securitisation Event 2026, market leaders explore how growth, regulation, new asset classes, AI, technology and data centres are reshaping Europe’s securitisation market. Expect concrete cases, practical insights and open debate across the full value chain. Join the largest annual Dutch securitisation platform and connect with the entire structured finance community.
Securitisation Event
The Securitisation Event, where more than 250 participants come together annually to discuss developments...
Download the brochure
Share
