Independent comparison of vendors building AI for issuer in-house collections, third-party collectors and debt-purchaser portfolios in 2026, with a focus on GDPR Article 22 (solely-automated decisions), CFPB Regulation F on debt-collection practices, the FCA Consumer Duty in the UK, and EU AI Act obligations where AI is used to evaluate creditworthiness or vulnerability [1][2].
Vendors were selected on five criteria: depth of EU AI Act practice (specifically Annex III point 5(b) where the workload touches creditworthiness assessment of natural persons), debt-collection-vertical specialisation evidenced by named issuer or collector deployments, a written delivery methodology, multilingual delivery across at least three European languages, and a citation chain in shipped outputs (account-level evidence linking each automated action to source data and policy).
The list is ordered by shape of fit, not ranked. The middle of the list is not worse than the top. Each vendor entry includes a tradeoff section because no single vendor fits every collections mandate.
Honesty disclosure: Impetora is one of the nine vendors below. We have written our own entry in the same factual register as the others. Where we make a comparison we cite the competitor's own website or a public registry. We do not invent statistics and we do not trash competitors.
Verification: every vendor below was confirmed operating as of April 2026 by checking their public website and a recent press mention.
Five practical tests separate serious debt-collection AI consultancies from generalists with a financial-services slide deck. First, can the vendor produce, in writing, a GDPR Article 22 design pattern showing how solely-automated decisions on affordability, hardship classification or vulnerability detection are bounded by meaningful human review [1]? Second, has the vendor read CFPB Regulation F (effective 30 November 2021) on contact-frequency caps, communication channels and validation notices, and can they articulate how AI behavioural triggers respect it [6]? Third, do shipped outputs include account-level evidence linking each automated action to source data, applicable policy and the human reviewer where required? Fourth, can the vendor demonstrate FCA Consumer Duty good-outcome evidence (UK) or the equivalent vulnerable-customer detection logic the relevant national regulator expects? Fifth, does the vendor name a senior engineer who will be hands-on rather than staffing a delivery pyramid?
The EBA guidelines on loan origination and monitoring, applicable since June 2021, also bind any AI used in renegotiation or pre-litigation hardship decisions for EU regulated lenders. Vendors that cannot cite Article 22, Regulation F, Consumer Duty and the AI Act in the proposal should not progress to shortlist.
The EU AI Act, Regulation (EU) 2024/1689, treats AI used to evaluate creditworthiness or credit score of natural persons as high-risk under Annex III point 5(b) [7]. Where collections AI feeds renegotiation, settlement or write-off decisions that materially affect a consumer's credit profile, the workload can be in scope. Article 43(2) carves a partial exception for AI inside an existing financial-services internal control framework, but the substantive obligations still apply.
GDPR Article 22 governs solely-automated decisions producing legal or similarly significant effects on the data subject. The CJEU SCHUFA judgment (C-634/21, December 2023) clarified that a credit-scoring output relied on by a downstream lender is itself an Article 22 decision [1]. The practical implication for collections AI is that any tool which automatically classifies an account as suitable for litigation, write-off, or hardship rejection without meaningful human input is likely an Article 22 decision and needs the corresponding safeguards.
CFPB Regulation F (12 CFR Part 1006), effective 30 November 2021, caps phone-call frequency at seven attempts in seven days per debt and requires opt-out for electronic communications [6]. AI behavioural triggers that select contact moments must be configurable so that the cap is enforced at the channel level. The FCA Consumer Duty (PS22/9, in force July 2023) requires UK firms to deliver good outcomes for retail customers including those in financial difficulty; AI vulnerability detection must produce evidence the firm can put in front of the regulator.
Buy when: the workload is digital-first consumer collections at volume on a generic US or UK consumer portfolio, and a productised platform such as TrueAccord already covers 80% of the use case with the regulator references and licensing the issuer needs. Productised economics beat custom builds where the issuer is happy to outsource the agency relationship.
Build when: the workload is specific to your book (specialist commercial debt, multilingual European pre-litigation, a particular regulator's vulnerability framework), where confidentiality requires a sub-processor footprint a SaaS vendor cannot offer, where the issuer wants to retain the agency licence and brand surface, or where AI Act and Article 22 conformity evidence must be owned by the buyer.
Hybrid when: digital outreach is bought from a productised platform but vulnerability detection, hardship classification and agent-call QA are custom because the issuer's policy taxonomy and tone of voice are bespoke. Most enterprise collections AI in 2026 ends up here.
Send the same five written questions to every vendor on the shortlist. The answers will sort the field faster than any analyst report.
Vendors that send back marketing PDFs in response to these questions are not ready. Vendors that respond with a redacted prior conformity assessment and a vulnerability-detection logic document are.
Ready to scope your project? Submit a short brief and we reply within one business day.
Tell us what you would like to build. We reply within one business day.