More case files, fewer people. That is the state of the sector. According to the KBvG annual report 2025, 2,161,070 debt-collection files were opened last year, 9.5% more than in 2024. At the same time, the number of FTEs in the sector fell to roughly 2,492, lower than in previous years, while around 3.5 million files were still being handled at year end. Staff costs make up the largest share of expenditure. Anyone in that position who hears that nearly half of the work can be automated wants to know one thing above all: which half exactly, and what changes in the firm as a result?
Which half: official acts or everything else?
The annual report itself gives the first clue. The KBvG writes: "The number of official acts provides insight into the legal part of the work, such as serving documents and levying attachment. A large part of the work also consists of contact with debtors, consultation with creditors and coordination with municipalities, administrators and other parties involved."
In 2025, 1,523,200 official acts were carried out, 5.3% fewer than in 2024. Under article 2 of the Dutch Bailiffs Act (Gerechtsdeurwaarderswet), those acts are the exclusive domain of the bailiff, with fees laid down in the Btag. Everything around them, the analysis, the correspondence, the coordination, is work that requires no official authority and that, according to the annual report, makes up a large part of the work.
Two external estimates point in the same direction. The independent website benikeenrobot.nl estimates, using its own task model (time spent multiplied by an automation score per task), that 46% of a bailiff's work can be automated. That is an indicative estimate, not scientific research, but the breakdown per task is instructive: drafting legal documents 70%, issuing summonses 57%, collecting debts 47%, executing attachments 40%, advising clients and debtors 23%. Goldman Sachs, in March 2023, put the share of tasks in legal occupations in the United States exposed to automation by generative AI at around 44%.
The KBvG position fits this picture. In her column "AI as bailiff: opportunity or threat?", chair Chris Bakhuis calls AI a tool, not a replacement. Automatable in her view: the financial analysis of the debtor, file analysis, communication with debtors and the detection of problematic debt. What remains human is "making an intuitive assessment of a situation" and "empathy and judgement during a home visit". Her summary: "AI is the future, but the human dimension remains the most important thing of all!"
What can actually be automated, per workflow
File analysis
With more than 2 million new debt-collection files a year, the first question is always the same: what is in the file, what is missing and what is the next step? Sorting and summarising documents, flagging missing data and drawing up a first financial picture of the debtor are tasks a language model handles well. The judgement whether the file is ready for an official act stays with the bailiff.
Checking draft writs and summonses
Drafting legal documents scores highest in the benikeenrobot.nl estimate, at 70%. In practice this is about checking drafts: are the party details, the deadlines, the references to the procedural rules and the cost items from the Btag correct? A second pair of eyes that does not tire, with the bailiff who signs.
Correspondence and payment arrangements
Standard letters, answers to frequently asked questions and a proposal for a payment arrangement lend themselves well to preparation. The tone is sensitive here, though: according to the KBvG, the sector recorded 1,881 complaints against firms and 651 reports of aggression in 2025, both lower than in 2024 (2,257 and 771). Every outgoing text therefore remains human work in the final check.
Case-law and disciplinary-law check
Before a less common step, it is useful to know what the Chamber for Bailiffs in Amsterdam, and on appeal the Amsterdam Court of Appeal, have ruled in comparable situations. That research is eminently automatable, provided every finding carries a source reference and the bailiff reads the ruling personally.
Attachment-free threshold and proportionality
Calculating and checking the attachment-free threshold (beslagvrije voet, article 475da of the Dutch Code of Civil Procedure) is rule-based arithmetic. The decision whether and how to levy attachment, and whether that is proportionate to the situation, is not. In 2025, according to the annual report, 181,039 attachments on periodic income were levied, 10.6% fewer than in 2024. Evictions rose instead: 10,443 announced and 3,649 carried out, respectively 20.3% and 22.8% more than in 2024. In those far-reaching steps the on-site assessment is decisive.
What cannot be automated
Three things stay with the human. The home visit, with the empathy and judgement Bakhuis refers to. The weighing around the duty to act (ministerieplicht): whether an instruction should be carried out in this form, or whether there is reason to refuse or to consult. And the proportionality assessment for far-reaching measures such as attachment and eviction. They are the tasks for which the authority and the disciplinary law exist.
Preconditions for the firm
Anyone deploying AI on the non-official part must arrange four things.
Source references. A statement without a citation is unusable in bailiff practice. Every outcome must be traceable to statute, regulation or ruling.
Human final judgement. The bailiff signs and is accountable under disciplinary law before the Chamber for Bailiffs. A tool does not change that.
Wki and registration. Firms that also offer debt-collection services fall under the Debt Collection Services Quality Act (Wki), with mandatory registration at Justis. The quality requirements apply just as much to correspondence prepared automatically.
GDPR. Files contain financial data and personal data of debtors. Where that data is processed and by whom must be settled in advance.
The sector is small and fragmented: 108 firms as of 1 January 2026, compared with 117 in 2024, and 89% of them are small or medium-sized, while the large firms generate two thirds of turnover, according to the KBvG annual report 2025. For a firm with a handful of bailiffs this is not an IT project but a choice of which workflow comes first. The annual report notes that digital facilities play an ever greater role in execution, with the Digital Attachment Register, Schuldenwijzer and the SNG data exchange. AI for the analysis and writing tasks is the next layer on that infrastructure.
What LEO does here
LEO, the AI colleague from PrudAI, knows seven professional roles, including the bailiff. In that role LEO reasons from the Bailiffs Act, the Btag, the Code of Civil Procedure (enforcement and attachment) and the KBvG regulation on professional and conduct rules. LEO works with the national procedural rules and the Beslagsyllabus, which are not in the ECLI database, and with two KBvG bailiff corpora: disciplinary tribunal rulings and enforcement doctrine. For disciplinary law LEO searches Tuchtrecht.nl live, including rulings on bailiffs. Every statement carries a source reference and the bailiff assesses the outcome personally.
Frequently asked questions about AI in bailiff practice
Does AI replace the bailiff?
No. Under article 2 of the Bailiffs Act, official acts are reserved to the bailiff, and the KBvG chair calls AI a tool and not a replacement. The home visit and the on-site assessment remain human.
Which tasks come first?
The tasks with the highest automation score in the benikeenrobot.nl estimate are drafting legal documents (70%) and issuing summonses (57%). In practice a firm usually starts with file analysis and checking drafts, because many hours sit there and the bailiff keeps the final check.
What about disciplinary law?
The bailiff remains accountable under disciplinary law before the Chamber for Bailiffs in Amsterdam, with appeal to the Amsterdam Court of Appeal. A tool does not change that responsibility. That is why a source reference with every outcome is a hard requirement.
May AI contact debtors directly?
Communication with debtors is automatable according to the KBvG chair, but the tone is sensitive and the firm remains accountable. The sensible approach is to let AI prepare texts and have a staff member assess the sending. For debt-collection services, the quality requirements of the Wki apply as well.
Is this only for large firms?
No. According to the KBvG annual report 2025, 89% of the 108 firms are small or medium-sized. That is precisely where every saved analysis and writing hour counts, and choosing one workflow as a starting point does not require a large IT project.