Fintech baropaper 2026
Returning research
For many finance professionals, AI feels like the next logical step in their processes. But adding a digital colleague where it truly counts is far from straightforward. Other sectors are experimenting freely, weaving AI deep into their core processes. Finance is moving more cautiously. Legislation is rightly strict, and it keeps the sector alert when it comes to AI. Or does it?
For the ninth year running, Onguard's FinTech Barometer surveyed 336 CFOs, finance managers and credit management professionals in the Netherlands. AI was on the agenda last year too, when it was mostly about the promise. This year, it's about the practice. How are finance professionals actually using AI? How do they stay in control, and how do they make sure decisions remain accountable? This research shows that AI adoption in the sector has outpaced control over it, and that shows up even at the top.
For the ninth year running, Onguard's FinTech Barometer surveyed 336 CFOs, finance managers and credit management professionals in the Netherlands. AI was on the agenda last year too, when it was mostly about the promise. This year, it's about the practice. How are finance professionals actually using AI? How do they stay in control, and how do they make sure decisions remain accountable? This research shows that AI adoption in the sector has outpaced control over it, and that shows up even at the top.
"Getting a grip on AI is a governance question, not an IT issue. If you're a CFO and you can't explain an AI decision, you've lost control over your own accountability. That's exactly why we keep running this research, year after year." – Dennis Kromhout van der Meer, managing director bij Onguard
In 30 seconds: what this report shows
EU AI Act
The EU AI Act has been in force since August 2024 and applies to every organisation in Europe that uses or offers AI services or tools. It sorts AI systems into four risk categories: minimal, limited, high and unacceptable. The higher the risk, the stricter the requirements.
One application matters most for finance and credit management: AI that assesses the creditworthiness or credit score of individuals. That falls squarely into the high-risk category. From December 2027, this means strict requirements around human oversight, logging, documentation and explainability towards the people affected.
AI on the rise
Chapter 1
AI is advancing quickly in finance. Usage right now is split almost fifty-fifty: 48% don't use it, 45% do, and 7% aren't sure. What stands out is that CFOs report using AI more than finance staff, which flips the original promise of AI on its head. The repetitive work AI has the most potential for, working through files, answering emails, chasing debtors, sits with the group least likely to say they use it.
"AI is something of a hype in our field. Usage is growing, but I'm not yet seeing it speed up the primary process or the heart of credit management. It's mostly being used on the edges, think ChatGPT or Claude alongside Excel. That doesn't mean organisations are behind, though. The moment an AI-based decision touches a customer's creditworthiness, you fall under the EU AI Act, and you need accountability and security in place before you even start. That's why we regularly see organisations asking, in tenders, whether AI features in software can be switched off or managed independently. These are companies that understand exactly what they're getting into. In finance, limited AI use is sometimes a deliberate choice."
– Willand Brienen, teamlead AI solutions & product bij Onguard
Ik maak gebruik van AI op m'n werk
55%
CFO's
50%
Financial management
36%
Finance employees
The explainability gap
Chapter 2
Using AI in finance means being able to explain, at any moment, how a decision came about, to the board, an auditor or a regulator. On many finance teams, that's not the reality. Only 18% can fully substantiate an AI-driven decision on lending or debt collection.
"Being able to mostly substantiate an AI-driven decision sounds reassuring, but it just as easily means you don't know what AI is doing behind the scenes for most of it. When you're granting a loan or collecting a debt, that's a real problem. How explainable an AI decision is depends on the type of AI you're using. Ask a language model to summarise an email, and you can't trace why it picked up one detail and left out another. Work with a scoring model, though, and there are rules and formulas behind it that you can document and justify. This matters well beyond compliance, too. Just look at the fallout when a government body used an algorithm it couldn't explain: the reputational damage was enormous. For financial institutions, the stakes are even higher. Reconstructing what a model did after the fact simply isn't possible. You have to build that in from the start." – Marco Bontje, international sales consultant bij Onguard
To what extent can you substantiate an AI-driven decision on granting a loan or collecting a debt to the board, an auditor or a regulator?
Speed versus control
Chapter 3
AI speeds up processes. But what does that do to the ability to keep those processes under control?
Across finance broadly, and in the collections process specifically, most professionals see AI keeping control at roughly the same level. The question is: the same as what? The previous chapter showed that fewer than one in five can fully substantiate an AI-driven decision. If most people can't, then control staying at that same level is a questionable achievement. At the same time, almost a third of professionals believe that AI's speed comes at the expense of control.
Across finance broadly, and in the collections process specifically, most professionals see AI keeping control at roughly the same level. The question is: the same as what? The previous chapter showed that fewer than one in five can fully substantiate an AI-driven decision. If most people can't, then control staying at that same level is a questionable achievement. At the same time, almost a third of professionals believe that AI's speed comes at the expense of control.
The effect of AI on speed and control within finance
"Where the balance between speed and control should sit depends entirely on where you apply AI. Sending payment reminders isn't especially high-stakes, so it needs little control. A major budgeting decision is a different story, one where speed matters far less. Given how sensitive our field is, speed almost always takes a back seat to quality. Cut corners there, and that speed gain comes back to bite you sooner or later." – Willand Brienen, teamlead AI solutions & product bij Onguard
The effect of AI on speed and control in the collections process
Human oversight and ownership
Chapter 4
Control, then, is never a given when it comes to using AI. In debt collection specifically, it mostly comes down to how the processes around the technology are set up.
According to finance professionals, the biggest bottleneck for AI is decisions that were never recorded, something AI can't factor in when automatically collecting payments. The same thread runs through the other three bottlenecks: a lack of structure.
According to finance professionals, the biggest bottleneck for AI is decisions that were never recorded, something AI can't factor in when automatically collecting payments. The same thread runs through the other three bottlenecks: a lack of structure.
The four biggest AI bottlenecks in collecting outstanding payments
34%
Decisions made over email or informal agreements
24%
Ad hoc manual workarounds
22%
Verouderde ERP-systemen die niet goed aansluiten
18%
Unclear process and data ownership
If AI can only build on what's been recorded, people have to catch the rest. Yet half of finance professionals say human oversight is often, though not always, part of their AI processes. Only one in three has it built in by default.
"AI simply makes decisions for you, whether it has all the relevant information or not. What matters most is that the information it does have is reliable and accessible, and that you can justify the outcomes afterwards, even when the system doesn't have the full picture. In collections, a lot of the important information doesn't live in the numbers. It lives in the nuance around them. Whether a debtor is struggling, for instance, is something you'll never read off the figures alone. That's exactly where the most essential part of our profession sits: the human factor and maintaining the customer relationship. And that's the time AI can, in theory, give you back." – Marco Bontje, international sales consultant bij Onguard
Human oversight is part of our AI processes
- Often
- Always
- Rarely
- Never
Actual human oversight
Almost seven in ten organisations regularly need to check or correct AI decisions. Human oversight, then, is being put to real use, not just promised. It's the group correcting AI-driven decisions regularly that deserves the most attention: for them, intervention has become structural. That ties straight back to where this chapter started. When the biggest obstacles for AI are unrecorded decisions and informal agreements, human correction of AI's work is less a surprise than a logical consequence. People catch what the system can't see.
So how does this play out specifically in collections?
So how does this play out specifically in collections?
Frequency of AI-driven decisions leading to corrections, exceptions or manual review
Collections as a case study
Chapter 5
The biggest risk of AI-driven collections within the organisation
Concerns about AI in collections
Concerns about AI in collections span several fronts, and professionals disagree on what the biggest risk actually is. One figure stands out: the 20% who flag a governance risk, insufficient grip at board level. That concern plays out in practice too. Over half (54%) admit they couldn't fully explain or trace an AI decision if an auditor asked today. Strikingly, CFOs are the most likely to admit this (74%), even though they're the ones ultimately accountable for these decisions.
That's not to say these AI-driven processes run without any human oversight. 63% regularly or occasionally step in themselves, correcting or reversing AI-driven decisions. So the real test is less about whether people step in, and more about whether that intervention holds up to scrutiny afterwards.
That's not to say these AI-driven processes run without any human oversight. 63% regularly or occasionally step in themselves, correcting or reversing AI-driven decisions. So the real test is less about whether people step in, and more about whether that intervention holds up to scrutiny afterwards.
"AI in collections is still in its infancy at many organisations. That said, some have already got this right. Every AI decision gets logged in an audit trail down to debtor level, including the data it was based on. AI offers a recommendation, a person makes the final call and signs off, and that step gets logged too. That way, an AI-driven decision always stays traceable."
– Dennis Kromhout van der Meer, managing director bij Onguard
How often do AI-driven collections decisions need to be corrected or reversed?
Control starts with a choice
Conclusie
AI adoption and control need to move at the same pace. The FinTech Barometer 2026 shows a sector integrating AI into its processes cautiously. Usage sits at almost exactly fifty-fifty.
That caution doesn't mean AI is being built on solid ground, though. Fully substantiating an AI-driven decision is the exception rather than the rule. Wherever AI speeds up a process, professionals see control slip in return. This is a structural problem: decisions and agreements still aren't always recorded, which keeps exactly the nuance and context around a process out of AI's view. Whether organisations realise it or not, every AI justification stays incomplete as long as that's the case. Human oversight has held things together so far, but not consistently. At most organisations, human intervention in AI decisions is common, though it's not a standard part of every process. That gap runs all the way to the top: it's the CFO, the person ultimately accountable, who is least able to explain an AI decision. Knowing what you decide and why, especially once AI is involved, comes down to arranging control and grip in advance, not once an auditor, a regulator, a customer or new legislation forces the question.
That caution doesn't mean AI is being built on solid ground, though. Fully substantiating an AI-driven decision is the exception rather than the rule. Wherever AI speeds up a process, professionals see control slip in return. This is a structural problem: decisions and agreements still aren't always recorded, which keeps exactly the nuance and context around a process out of AI's view. Whether organisations realise it or not, every AI justification stays incomplete as long as that's the case. Human oversight has held things together so far, but not consistently. At most organisations, human intervention in AI decisions is common, though it's not a standard part of every process. That gap runs all the way to the top: it's the CFO, the person ultimately accountable, who is least able to explain an AI decision. Knowing what you decide and why, especially once AI is involved, comes down to arranging control and grip in advance, not once an auditor, a regulator, a customer or new legislation forces the question.
"For Onguard, the coming years are less about maximising AI and more about using it deliberately. Not every process calls for the same degree of autonomy, control or explainability. Sending a payment reminder is a different matter from deciding to send a customer to collections. AI can add value in both situations, just in different ways. That choice is exactly where the maturity lies. Where do you let AI support, advise or act independently, and what level of control fits each of those? That's how we approach further AI adoption at Onguard, too." -Willand Brienen, Team Lead AI Solutions & Product at Onguard
What the data tells us
Best practices:
- Know where AI can and can't help Not every process is suited to automation. AI is good at flagging risks early, negative news about a customer, for example. But the next step, how you respond to the customer, remains work for people. Rather than rushing to be first, wait until you're confident AI can handle it, especially anywhere the stakes are higher, such as the customer relationship. Decide, process step by process step, whether AI takes over a task or only supports it. Sometimes, less AI is the better call.
- Build in explainability upfront, not as an afterthought Reconstructing what an AI model did after the fact simply doesn't work. So record, for every decision, which data was used, what AI recommended and who signed off on it. Skip these steps, and AI stays a black box you can't fall back on the moment a key stakeholder comes asking.
- Innovate with AI where it counts The biggest wins from AI often come from small things rather than the most ambitious idea. Sorting incoming emails, running analyses or drafting recommendations can already be fully automated today. These applications might not be the most impressive, but they're immediately usable and explainable. And that's what counts in the end.
About the research
Every year, Onguard researches how finance professionals view the financial trends and developments shaping the moment, from new technologies to new payment methods. This research, the FinTech Barometer, has now been conducted for the ninth year running. 319 finance professionals took part, 101 of them CFOs and 218 finance professionals.
About Onguard
Onguard speeds up payments and improves cash flow for businesses through smart credit management. Our solutions give you insight into payment behaviour, help you minimise financial risk and sharpen your communication around unpaid invoices. The result? Faster payments, fewer write-offs and lower customer churn. With more than 30 years of experience, we help optimise working capital and free up valuable time with your customers by handling the day-to-day. The result: get paid quicker, DSO cut by up to 30%, and collection costs down 15 to 30%.
Meer weten?
Got questions about our solution, or how to set up credit management as effectively as possible within your organisation? Get in touch; we're happy to help.
Call us:
+31 (0) 88 256 66 66
+44 (0) 20 396 683 24
E-mail us:
contact@onguard.com
Copyright © Onguard. All rights reserved
The information in this baropaper has been compiled with care[cite: 1]. Nevertheless, Onguard disclaims all liability for any consequences resulting from incomplete or inaccurate information contained within this baropaper.