AI for debt collection, from portfolio scoring to compliance audit trail.

AI used for creditworthiness assessment and credit scoring is classified as high-risk under EU AI Act Annex III §5. Whether your specific recovery workflow is high-risk depends on whether it produces or materially informs a decision about a natural person's creditworthiness, payment terms, or settlement eligibility. We classify the system in the discovery phase: if it is high-risk, we ship the full conformity-assessment scaffolding (risk management, data governance, technical documentation, record-keeping, transparency, human oversight, accuracy, robustness, cybersecurity); if it is limited-risk we ship the proportionate controls. We never default to less than the regulation requires.

Article 22 prohibits decisions producing legal or similarly significant effects from being made solely on automated processing without explicit safeguards. In practice that means a documented human-in-the-loop step for any output that affects a debtor's payment terms, settlement eligibility, or escalation path. The AI surfaces a recommendation with citations to the source signals; a trained human approves or rejects it. We also build the right-to-contest mechanism: a debtor can request human review and the system produces the source-signal pack the reviewer needs in under one minute.

Outbound telephony is not Impetora's lane. Some debt collection deployments include phone-channel automation via specialised vendors (e.g. Ainora.lt for the Lithuanian market and a number of established platforms in the US and UK), and we are happy to advise on integration, but our consultancy work concentrates on the document, decision-support, and compliance audit layers: portfolio segmentation, payment-plan proposals, hardship flagging, adverse-action notices, recoveries forecasting. If a phone channel is the only thing you need, we will tell you so and point you at a vendor.

Vulnerability detection is a first-class use case, not a footnote. The system reads inbound correspondence, call transcripts, and complaint records, and surfaces signals that map to recognised vulnerability frameworks (financial hardship, bereavement, ill-health, mental-capacity concerns, language barriers). Every flag carries a citation to the source text. The downstream action is always human routing to a trained vulnerability specialist; the AI does not make the call on whether to forbear or to continue collections. The audit log captures the flag, the routing, the human decision, and the timing, which is what supervisors and ombudsmen ask for.

Yes. The delivery layer is built around your system of record, not the other way around. We ship integrations with TallyMan, Latitude by Genesys, Qualco, FICO Debt Manager, and the major in-house cores via documented APIs or queue-based bridges with idempotent writes and a manual reconciliation interface. The audit log writes regardless of where the data lands, so lineage is provable even when the downstream system is older.

A first engagement targets one workflow with a measurable baseline, runs 4 to 12 weeks to production, and lands as a single signed-off system inside one platform surface. Common scopes are: portfolio segmentation across one product line; payment-plan proposal automation across one customer segment; vulnerability flagging across all inbound channels; or compliance audit-trail automation across the full recovery workflow. Submit a project with the workflow you have in mind and the rough portfolio size, and we scope the discovery phase before any code is written.

By default, all processing and storage runs in EU regions on infrastructure under EU jurisdiction. We support regional pinning when a regulator or contract requires it. Original account data lands in immutable EU object storage with hashes recorded in the audit log. We do not train any model on your debtor data, full stop. If a use case requires US-resident processing for a US-only portfolio, we expose that as an explicit configuration toggle, never a default.

Pricing depends on scope and regulatory complexity. We do not publish a flat rate because the scope variation across debt collection AI is wide: a single-segment propensity model on a uniform portfolio is a different build from a vulnerability-flagging stack across all inbound channels with full conformity-assessment documentation. Submit a project with the workflow and rough portfolio size, and we come back with a discovery proposal within one business day - see the intake form for budget bands.