[{"data":1,"prerenderedAt":701},["ShallowReactive",2],{"uc-propensity-to-pay-scoring":3,"uc-regulations":480},{"useCase":4,"evidence":204,"blitsAiDeployments":337,"benchmarks":338,"indicative":358,"related":361,"indexability":478,"includeUnpublished":210},{"title":5,"shortTitle":6,"seoTitle":7,"metaDescription":8,"status":9,"definition":10,"aliases":11,"industries":17,"functions":19,"patterns":22,"channels":25,"audience":28,"autonomy":29,"adoptionStage":30,"problem":31,"problemStats":32,"howItWorks":38,"valueDrivers":39,"kpis":44,"indicativeValue":51,"macroEstimates":84,"feasibility":85,"implementation":98,"risk":144,"blitsAi":181,"faq":183,"related":193,"datePublished":199,"dateModified":199,"lastVerified":199,"changelog":200,"slug":203},"AI propensity to pay scoring for arrears and debt collection","Propensity to pay scoring","Propensity to pay scoring with machine learning","Propensity to pay models score debtors so collectors choose whom to contact and how. HMRC scores about 4 million debtors a month, DWP 100,000 paying parents.","published","Propensity to pay scoring is a machine learning model that estimates how likely each debtor is to pay, or to miss an upcoming payment, so collectors can pick the cases and contact approach. The score orders worklists and selects the letters, calls or support a debtor receives, while people still decide on payment plans, enforcement and write offs.",[12,13,14,15,16],"propensity to pay model","likelihood to pay scoring","arrears risk scoring","collections prioritisation","payment risk scoring",[18],"government",[20,21],"collections-and-recovery","risk-management",[23,24],"prediction-and-scoring","classification-and-routing",[26,27],"internal-tools","api","back-office","assist","early-adopters","Most collections teams hold more overdue accounts than they have collectors. Without a score, the\nworklist is ordered by balance, by age of the debt or by the order in which cases arrived, and every\ndebtor in the same bucket receives the same letters and calls. That spends effort on people who would\nhave paid after one reminder and misses people whose payments are slowly falling away. HMRC's\nBusiness Propensity to Pay model replaced a simple segmentation by exposure to the pandemic, and\nbefore that segmentation most customers received the same debt collection journey.\nCamden Council says its rent teams struggled to focus on the right areas and missed chances to give\npreventive support to tenants who could not pay.\n\nA score also separates two groups who need opposite treatment. A debtor who is likely to pay needs a\nprompt and an easy way to pay. A debtor whose payments are declining may be in financial difficulty\nand needs an earlier and more careful conversation or a referral for support. Used carelessly, the\nsame score can do harm. Australia's Robodebt scheme shows what happens when an automated process raises\ndebts, puts the burden of disproving them on the person and has no human check.",[33],{"statement":34,"sourceTitle":35,"sourceUrl":36,"year":37},"Camden Council reports that its total council housing rent arrears were second highest in London and nearing 18 million pounds in December 2023.","Camden Council: RentSense (algorithmic transparency record)","https:\u002F\u002Fwww.gov.uk\u002Falgorithmic-transparency-records\u002Fcamden-council-rentsense",2023,"1. **Build the account history.** Link every debt a customer holds, as HMRC does across VAT, PAYE and\n   Corporation Tax, and add payment history, balances, changes of payment method and, where it is\n   lawful, contact history.\n2. **Train and calibrate the model.** A supervised model, usually gradient boosted trees (HMRC and DWP\n   both use them), predicts whether a debtor will pay or engage within a set window. HMRC calibrates the\n   raw output with a logistic regression so the score reads as a likelihood.\n3. **Band the scores.** Scores become bands such as high, medium and low, because collections systems\n   and caseworkers act on bands.\n4. **Choose the treatment.** The band, together with other rules, sets the timing, frequency, channel\n   and wording of the first contact, or places the case on a caseworker's list with the reasons and\n   suggested actions. HMRC says the score is only one factor in the collection path.\n5. **Leave consequences to people.** The score orders work and selects communications. People agree\n   payment plans, handle vulnerability and decide on enforcement.\n6. **Retrain and monitor.** DWP trains a new model every month and accepts it only above a precision\n   floor. HMRC runs its model monthly, watches input drift on a dashboard and reviews it every year.",[40,41,42,43],"risk-reduction","employee-productivity","cost-to-serve","customer-experience",[45,46,47,48,49,50],"recovery-rate-uplift","revenue-recovered","accuracy","cost-savings","interactions-handled","delinquency-rate-reduction",{"referenceOrg":52,"inputs":53,"formula":79,"currency":80,"period":81,"resultLabel":82,"caveat":83},"A water utility with 1 million household accounts",[54,59,66,72],{"key":55,"label":56,"low":57,"high":57,"unit":55,"note":58},"accounts","Household accounts",1000000,"The reference utility.",{"key":60,"label":61,"low":62,"high":63,"unit":64,"note":65},"arrearsShare","Share of accounts that fall into arrears in a year",0.08,0.12,"fraction of accounts","Editorial assumption, replace with your own arrears data.",{"key":67,"label":68,"low":69,"high":70,"unit":71,"note":65},"overdueBalance","Average overdue balance per account in arrears",150,300,"USD per account",{"key":73,"label":74,"low":75,"high":76,"unit":77,"note":78},"recoveryUplift","Relative increase in overdue money recovered after targeting",0.02,0.06,"fraction of overdue balance","Deliberately below the 8 percent total growth in collections that the New York State tax department reports for its first year, because the authors attribute only 40 to 50 million of the 83 million dollar growth to the system and no other record on this page publishes a recovery result.","accounts * arrearsShare * overdueBalance * recoveryUplift","USD","per year","Additional overdue money recovered","Gross cash only. It leaves out the cost of building and running the model, money that would have been recovered anyway, the effect of earlier support on bad debt and customer retention, and any cost of treating vulnerable customers differently.",[],{"complexity":86,"complexityNote":87,"dataPrerequisites":88,"integrations":93},"medium","The build needs an extract of account level payment history with known outcomes, usually a year or more, joined across every product a debtor holds, and a collections or case system that accepts a score and a band. HMRC tuned its model on 9.5 million rows and DWP trained on about 200,000 labelled cases. Wiring a score into letter timing or worklist order is a small change to a collections system. A score that sits near enforcement, disconnection or benefit decisions needs model risk review, bias testing and a recorded human decision.",[89,90,91,92],"Account level payment history with outcomes (paid, partly paid, broken arrangement) for at least a year","Debts linked across products or tax types at customer level","Contact and arrangement history from the collections system","A lawful basis and a record of which fields are excluded, such as protected characteristics",[94,95,96,97],"Collections or debt management system that can store a score and a band","Case management or caseworker desktop to show the score with reasons","Letter, SMS and email engine that can vary timing and wording by band","Data warehouse or feature store with monthly refresh and drift monitoring",{"steps":99,"guardrails":118,"humanInTheLoop":124,"kpisToInstrument":125,"failureModes":131},[100,103,106,109,112,115],{"title":101,"detail":102},"Name the outcome the model predicts","Pick one measurable event, such as a payment within 30 days of the first reminder, engagement with the recovery process (HMRC) or a missed payment on a case that is currently compliant (DWP). A vague target produces a score nobody can check.",{"title":104,"detail":105},"Decide what changes by band before training","Write down the letter timing, channel, caseworker order or support offer that differs for each band. A score that changes no action cannot show any value.",{"title":107,"detail":108},"Choose an explainable model and test on later months","Compare a few simple model families on the same data, as HMRC did with a single tree, a random forest and gradient boosting, and keep the most recent months aside as a hold out. Set the minimum precision a retrained model must reach, as DWP does.",{"title":110,"detail":111},"Test for bias before anyone uses the output","Keep protected characteristics out of the inputs, then compare who lands in each band against the equalities data you hold. Camden planned this check for its pilot and kept that data away from the supplier.",{"title":113,"detail":114},"Measure uplift against a holdout","Leave a random share of debtors on the old treatment for a few cycles and compare the money recovered, the cure rate and the cost to collect in both groups. The difference is the uplift the model can claim.",{"title":116,"detail":117},"Monitor monthly and review every year","Watch input drift, band mix and complaint rates each month, and retrain when behaviour changes. HMRC's current model superseded an earlier segmentation built around exposure to the pandemic.",[119,120,121,122,123],"The score orders work and selects communications; no enforcement, write off or disconnection follows from a score alone","Protected characteristics and special category data stay out of the model inputs and are used only to test outputs","A hardship or vulnerability flag overrides the score and routes the case to a person","Caseworkers see the reasons for a score and the suggested actions beside the number","A retrained model goes live only above a precision floor and after a drift check","Collectors and caseworkers read the score, check the case file and decide on contact, payment plans and referrals for support. Managers sample low band cases every month to catch debtors the model ranked too low, and review the model owner's bias analysis before each retraining.",[126,127,128,129,130],"Payment or cure rate by band, against a holdout group on the old treatment","Precision of the high risk band, meaning the share of high risk cases that did miss a payment","Contacts and complaints per 1,000 debtors by band","Share of debtors with a vulnerability flag in each band","Cost to collect per pound recovered",[132,135,138,141],{"title":133,"detail":134},"The score is treated as fact","Officers read a probability as a certainty. Camden listed this risk in its record. Train staff on the limits, show reasons beside the score and track how often officers override it.",{"title":136,"detail":137},"Vulnerable debtors fall into the low band","A debtor whose payments are steady until a crisis hits looks safe. Camden also lists the risk of high priority cases treated as low priority. Add vulnerability signals and sample low band cases.",{"title":139,"detail":140},"The model finds people who would have paid anyway","High scores can simply mark customers who pay after one reminder. Measure uplift against a holdout group and report it next to accuracy.",{"title":142,"detail":143},"Behaviour changes and the model does not","A model trained before a price rise, benefit change or pandemic stops describing today's debtors. Monthly drift checks and scheduled retraining catch it.",{"euAiAct":145,"regulations":148,"guidance":153,"controls":170,"incidents":176},{"tier":146,"basis":147},"context-dependent","A score that only orders a collector's worklist or chooses the timing of a reminder is not listed in Annex III. The same score becomes high risk under Annex III point 5(b) if it is used to evaluate the creditworthiness of natural persons, and may fall under point 5(a) if a public authority uses it to grant, reduce, revoke or reclaim essential public benefits or services. Keep the scope to prioritisation and document where the line sits.",[149,150,151,152],"eu-ai-act","gdpr","uk-gdpr","uk-atrs",[154,160,165],{"title":155,"issuer":156,"region":157,"url":158,"note":159},"Guidance on AI and data protection","Information Commissioner's Office","europe","https:\u002F\u002Fico.org.uk\u002Ffor-organisations\u002Fuk-gdpr-guidance-and-resources\u002Fartificial-intelligence\u002Fguidance-on-ai-and-data-protection\u002F","Covers fairness, accuracy and explanation duties when personal data feeds an AI system, including statistical accuracy of the scores.",{"title":161,"issuer":162,"region":157,"url":163,"note":164},"Guidance for firms on the fair treatment of vulnerable customers (FG21\u002F1)","Financial Conduct Authority","https:\u002F\u002Fwww.fca.org.uk\u002Fpublications\u002Ffinalised-guidance\u002Fguidance-firms-fair-treatment-vulnerable-customers","Sets out how firms should recognise and treat customers in vulnerable circumstances, which applies to FCA regulated firms that use a score to select collections treatment.",{"title":166,"issuer":167,"region":157,"url":168,"note":169},"Algorithmic Transparency Recording Standard hub","UK Government","https:\u002F\u002Fwww.gov.uk\u002Fgovernment\u002Fcollections\u002Falgorithmic-transparency-recording-standard-hub","The public records from HMRC, DWP and Camden Council on this page follow this standard and show what a public body discloses about a debt model.",[171,172,173,174,175],"Model card with the outcome predicted, training window, inputs, exclusions and the validation result","Bias and fairness analysis before launch and after each retraining, with a named owner","Monthly dashboard of drift, band mix, overrides and complaints","Logged score, band and treatment for every case, so a complaint can be traced","Public transparency record where the deploying body is a public authority",[177],{"title":178,"url":179,"note":180},"Royal Commission into the Robodebt Scheme: report volume 1 (archived)","https:\u002F\u002Fweb.archive.org\u002Fweb\u002F20241231202325\u002Fhttps:\u002F\u002Frobodebt.royalcommission.gov.au\u002Fsystem\u002Ffiles\u002F2023-07\u002Frobodebt_report_volume_1.pdf","Robodebt raised welfare debts by averaging income and placed the burden on recipients to disprove them. It used no propensity model. It is the best known example of an automated debt process without a human check, and the lesson applies to any score that triggers recovery action.",{"howToBuild":182},"The organization trains the scoring model on its own data platform, and the model writes a score, a band\nand the main reasons back to the collections or case system. Blits.ai reads the score and uses it in\nagents and flows. An **agent** for collectors and caseworkers uses **custom functions** (REST calls or SQL\nqueries) and **SQL knowledge bases** to read the caseload, explain why a case sits in its band and\ndraft the next contact for a person to review, on an internal channel such as **web chat or Microsoft\nTeams**.\n\nFor debtor contact, a **flow** picks the reminder wording and channel by band and sends it by\n**SMS, WhatsApp, email or voice**. **Guardrails** and **PII masking** protect the messages, and a hardship\nsignal triggers **human handover** with the conversation history. Any action above a set threshold,\nsuch as a payment arrangement, goes through an **agentic workflow** with **human in the loop**\napproval. **Test suites** replay realistic debtor conversations on every change, **monitors** and\n**analytics** follow band mix, contacts and handovers, and **model agnostic routing** with **EU and UAE\ndata residency** lets the organization choose where the language models run.",[184,187,190],{"question":185,"answer":186},"How accurate are propensity to pay models?","The published figures come from validation data. DWP reports accuracy of 0.86 and precision of 0.79 for its child maintenance model, which predicts missed payments on compliant cases. Mobysoft, the supplier of RentSense, claims typical accuracy of 95% or more, and Camden Council quotes that claim in its transparency record without verifying it. Accuracy depends on how many debtors miss payments, so compare precision and uplift against a holdout.",{"question":188,"answer":189},"Does propensity to pay scoring increase recovery?","No record on this page measures the recovery effect of a propensity to pay score. The closest result comes from New York State's tax department, whose system chooses collection actions for each delinquent taxpayer and is a related optimization method, not a payment likelihood score. Its collections rose 83 million dollars (8 percent) from 2009 to 2010, and the authors estimate that 40 to 50 million dollars of that came from the system, because collections normally grow 2 to 4 percent a year. HMRC, DWP and Camden Council disclose how their models work and how many cases they score, but no recovery result.",{"question":191,"answer":192},"Is propensity to pay scoring high risk under the EU AI Act?","It depends on use. Ordering a worklist or timing a reminder is not an Annex III use. Scoring that evaluates the creditworthiness of natural persons is high risk under point 5(b), and a public authority using a score to reclaim essential benefits or services may fall under point 5(a).",[194,195,196,197,198],"collections-and-hardship-agent","outbound-notice-drafting","credit-early-warning-monitoring","tax-compliance-risk-scoring","benefit-fraud-and-error-detection","2026-10-13",[201],{"date":199,"note":202},"First published","propensity-to-pay-scoring",[205,243,281,300],{"title":206,"useCases":207,"organization":208,"vendors":213,"summary":217,"stage":218,"year":219,"channels":220,"languages":221,"metrics":223,"outcomeDisclosed":232,"sources":233,"verification":238,"grade":240,"id":241,"organizationSlug":242},"HM Revenue and Customs: Business Propensity to Pay model for tax debt collection journeys",[203],{"name":209,"anonymized":210,"country":211,"region":157,"industry":18,"wikidata":212},"HM Revenue and Customs",false,"GB","Q166559",[214],{"name":215,"role":216},"In house (HMRC Debt Management, SAS Viya)","in-house","HMRC's Debt Management function runs a machine learning model called Business Propensity to Pay that scores about 4 million business debtors each month on how likely they are to engage with debt recovery. A gradient boosted decision tree, built in SAS Viya, learns from debt value and past repayment behaviour across VAT, PAYE and Corporation Tax debts linked at customer level. The segmented score helps decide which collection journey a debtor receives, including the timing, frequency and content of the first letters. HMRC says the model replaced a simple segmentation based on exposure to the pandemic (before that segmentation, most customers received the same journey) and that the score is only one factor in the journey. The record discloses no recovery results.","scaled",2026,[26],[222],"en",[224],{"kpi":49,"value":225,"unit":226,"qualifier":227,"period":228,"claimant":229,"quote":230,"sourceUrl":231},4000000,"count","approximately","per month","organization","The model is run on a monthly basis and creates scores for approximately 4 million customers each month.","https:\u002F\u002Fwww.gov.uk\u002Falgorithmic-transparency-records\u002Fhmrc-business-propensity-to-pay",true,[234],{"url":231,"title":235,"publisher":236,"date":237},"HMRC: Business Propensity to Pay (algorithmic transparency record)","GOV.UK","2026-09-09",{"level":239,"checkedAt":199},"source-verified","B","hmrc-business-propensity-to-pay","hm-revenue-and-customs",{"title":244,"useCases":245,"organization":246,"vendors":249,"summary":256,"stage":257,"year":258,"channels":259,"languages":260,"metrics":261,"outcomeDisclosed":232,"sources":274,"verification":278,"grade":240,"id":279,"organizationSlug":280},"DWP Child Maintenance Service: predicting payment breakdown on compliant cases",[203],{"name":247,"anonymized":210,"country":211,"region":157,"industry":18,"wikidata":248},"Department for Work and Pensions","Q1191394",[250,253],{"name":251,"role":252},"Tata Consultancy Services","integrator",{"name":254,"role":255},"Amazon Web Services (SageMaker)","platform","The Child Maintenance Service (CMS) in the Department for Work and Pensions (DWP) uses a machine learning model to predict which paying parents on compliant Collect and Pay cases are likely to miss a payment. A model trained each month on case, payment, liability and contact data (an XGBoost classifier with SHAP explanations) scores around 100,000 cases and shows caseworkers a high, medium or low risk with the reasons and suggested actions, so they can step in before a payment breaks down. Caseworkers decide what to do; the score is one input. The tool went to a wider group of caseworkers in March 2025 and DWP reports validation accuracy of 0.86 and precision of 0.79.","production",2025,[26],[222],[262,267],{"kpi":49,"value":263,"unit":226,"qualifier":227,"period":264,"claimant":229,"quote":265,"sourceUrl":266},100000,"cases scored each month","The tool is used at the start of each month to predict the risk of non-compliance, around 100,000 cases are processed by the model and presented to the CMS2012 System.","https:\u002F\u002Fwww.gov.uk\u002Falgorithmic-transparency-records\u002Fdwp-child-maintenance-service-cms-predictive-analytics-compliance",{"kpi":47,"value":268,"unit":269,"qualifier":270,"period":271,"baseline":272,"claimant":229,"quote":273,"sourceUrl":266},86,"percent","exact","validation data from May 2025","The model is accepted only when precision on validation data is at least 0.62; the record reports precision of 0.79.","Accuracy: 0.86 Precision: 0.79",[275],{"url":266,"title":276,"publisher":236,"date":277},"DWP: Child Maintenance Service (CMS) Predictive Analytics Compliance (algorithmic transparency record)","2025-10-30",{"level":239,"checkedAt":199},"dwp-child-maintenance-payment-breakdown-prediction","department-for-work-and-pensions",{"title":282,"useCases":283,"organization":284,"vendors":286,"summary":289,"stage":290,"year":258,"channels":291,"languages":292,"metrics":293,"outcomeDisclosed":210,"sources":294,"verification":297,"grade":240,"id":298,"organizationSlug":299},"Camden Council: RentSense pilot to prioritise council rent arrears cases",[203],{"name":285,"anonymized":210,"country":211,"region":157,"industry":18},"London Borough of Camden",[287],{"name":288,"role":255},"Mobysoft","Camden Council's housing income team piloted Mobysoft's RentSense for six months to decide which council tenants in rent arrears housing officers should contact first. The supplier's supervised machine learning reads rent transaction data from the council's NEC Housing system, looks at payment patterns and payment types such as direct debit and housing benefit, and predicts who will or will not pay. Each week officers get a prioritised list with plain reasons, such as an increasing balance or a broken arrangement. Mobysoft's own statement, quoted inside the record, is that accuracy is typically 95%+ across social landlords; Camden does not report it for the pilot. Camden says there is no automated decision making, that the supplier would not disclose how the algorithms weigh features, and that it would test the outputs for bias against equalities data held by the council. Total council rent arrears were nearing 18 million pounds in December 2023. No pilot result is published.","pilot",[26],[222],[],[295],{"url":36,"title":35,"publisher":236,"date":296},"2025-01-28",{"level":239,"checkedAt":199},"camden-council-rentsense-arrears-prioritisation",null,{"title":301,"useCases":302,"organization":303,"vendors":308,"summary":313,"stage":257,"year":314,"channels":315,"languages":316,"metrics":317,"outcomeDisclosed":232,"sources":326,"verification":335,"grade":240,"id":336,"organizationSlug":299},"New York State Department of Taxation and Finance: constrained reinforcement learning for delinquent tax collections",[203],{"name":304,"anonymized":210,"country":305,"region":306,"industry":18,"wikidata":307},"New York State Department of Taxation and Finance","US","north-america","Q7014085",[309,311],{"name":310,"role":252},"IBM Research",{"name":312,"role":252},"IBM Global Business Services","New York State's tax department built a collections system with IBM Research and IBM Global Business Services that decides, for each delinquent taxpayer, which collection action to take next. It models every taxpayer from about 200 features of their history and uses constrained reinforcement learning to choose actions that maximise long term returns within staff capacity and legal limits, instead of applying one rule to every case. The system became operational in December 2009. According to the department's own paper, total collections from delinquent revenue rose 83 million dollars (8 percent) from 2009 to 2010 with the same resources. Because collections had grown 2 to 4 percent a year before, the authors place the increase due to the system in its first year at 40 to 50 million dollars. The system chooses collection actions and is not described as a payment likelihood score. It is an early and well documented example, published in the Franz Edelman Award section of the INFORMS journal Interfaces; its current status is unknown.",2010,[26],[222],[318],{"kpi":46,"value":319,"unit":320,"currency":80,"qualifier":321,"period":322,"baseline":323,"claimant":229,"quote":324,"sourceUrl":325},50000000,"currency","up-to","first year after launch (2010), the authors' estimate of the increase due to the system, given as a range of 40 to 50 million dollars","Total collections grew 83 million dollars (8.22 percent) from 2009 to 2010; collections had grown 2 to 4 percent a year in earlier years.","increases ranging between 2–4 percent during the past several years; this places the revenue increase from CISS in its first year in the range of $40–$50 million already realized, with a promise of $120–$150 million over the three-year period originally projected.","https:\u002F\u002Fdmkd.cs.vt.edu\u002Fpapers\u002FInterfaces12.pdf",[327,331],{"url":325,"title":328,"publisher":329,"date":330},"Tax Collections Optimization for New York State (Interfaces, January to February 2012)","INFORMS Interfaces (author copy)","2012-02-01",{"url":332,"title":333,"publisher":334},"https:\u002F\u002Fwww.informationweek.com\u002Fit-sectors\u002Fibm-software-helps-ny-state-recover-83-million","IBM Software Helps NY State Recover $83 Million","InformationWeek",{"level":239,"checkedAt":199},"new-york-state-tax-collections-optimizer",0,[339,347,353],{"kpi":49,"label":340,"unit":226,"aggregate":210,"higherIsBetter":232,"n":341,"nUpTo":337,"median":342,"min":263,"max":225,"byClaimant":343,"vendorOnly":210,"points":344},"Interactions handled",2,2050000,{"organization":341,"vendor":337,"regulator":337,"independent":337},[345,346],{"evidenceId":241,"organization":209,"value":225,"qualifier":227,"claimant":229,"grade":240,"pooled":232},{"evidenceId":279,"organization":247,"value":263,"qualifier":227,"claimant":229,"grade":240,"pooled":232},{"kpi":47,"label":348,"unit":269,"aggregate":232,"higherIsBetter":232,"n":349,"nUpTo":337,"median":268,"min":268,"max":268,"byClaimant":350,"vendorOnly":210,"points":351},"Accuracy",1,{"organization":349,"vendor":337,"regulator":337,"independent":337},[352],{"evidenceId":279,"organization":247,"value":268,"qualifier":270,"claimant":229,"grade":240,"pooled":232},{"kpi":46,"label":354,"unit":320,"currency":80,"aggregate":210,"higherIsBetter":232,"n":337,"nUpTo":349,"median":299,"min":299,"max":299,"byClaimant":355,"vendorOnly":210,"points":356},"Revenue recovered",{"organization":337,"vendor":337,"regulator":337,"independent":337},[357],{"evidenceId":336,"organization":304,"value":319,"qualifier":321,"claimant":229,"grade":240,"pooled":210},{"low":359,"high":360},240000,2160000,[362,398,429,445,460],{"slug":194,"title":363,"shortTitle":364,"definition":365,"status":9,"industries":366,"functions":374,"patterns":376,"audience":380,"autonomy":381,"adoptionStage":30,"segment":382,"evidenceCount":383,"publicEvidenceCount":384,"organizations":385,"bestGrade":390,"headline":391,"lastVerified":397,"indexable":232},"AI agent for early collections and hardship support","Collections and hardship agent","A voice and messaging agent that contacts customers in early arrears to collect payments. It answers their inbound calls too, takes payments and sets up payment arrangements within preapproved rules, and recognises signs of hardship or vulnerability so those customers go straight to a trained person.",[367,368,369,370,371,372,373],"cross-industry","banking","payments","telecommunications","energy-and-utilities","automotive","professional-services",[20,375],"customer-service",[377,378,379,24],"voice-agent","conversational-agent","agentic-workflow","customer-facing","supervised-agent","lending",5,4,[386,387,388,389],"Day Knight & Associates","OTP Bank Romania","SameDay Auto Finance","Southwest Recovery Services","C",{"kpi":392,"label":393,"unit":269,"n":341,"nUpTo":337,"kind":394,"value":395,"qualifier":270,"claimant":396,"organization":388,"vendorReported":232},"cost-reduction","Cost reduction","reported",75,"vendor","2026-09-27",{"slug":195,"title":399,"shortTitle":400,"definition":401,"status":9,"industries":402,"functions":406,"patterns":410,"audience":414,"autonomy":415,"adoptionStage":30,"segment":28,"evidenceCount":416,"publicEvidenceCount":416,"organizations":417,"bestGrade":240,"headline":423,"lastVerified":428,"indexable":232},"AI for drafting customer letters and outbound notices","Outbound notice drafting","AI for outbound notice drafting writes the letters and notices operations must send at scale, such as arrears notices or decline letters. It also covers complaint responses, servicing confirmations and product change notices, from case data and approved templates and clauses, in the customer's language, for a person to approve where the notice is regulated.",[367,368,403,18,404,405],"insurance","healthcare","wealth-and-asset-management",[407,375,20,408,409],"operations","regulatory-compliance","claims",[411,412,413],"content-generation","rag-knowledge-assistant","translation","employee-facing","copilot",6,[418,419,420,421,422],"Acentra Health","Hiscox","Health Resources and Services Administration","Nationwide Building Society","SS&C Technologies",{"kpi":424,"label":425,"unit":269,"n":349,"nUpTo":337,"kind":394,"value":426,"qualifier":270,"claimant":396,"organization":427,"vendorReported":232},"processing-time-reduction","Cycle time reduction",25,"SS&C GIDS and RS","2026-09-26",{"slug":196,"title":430,"shortTitle":431,"definition":432,"status":9,"industries":433,"functions":434,"patterns":436,"audience":414,"autonomy":29,"adoptionStage":30,"segment":382,"evidenceCount":383,"publicEvidenceCount":384,"organizations":440,"bestGrade":240,"headline":299,"lastVerified":397,"indexable":232},"AI early warning and covenant monitoring for loan portfolios","Credit early warning and covenants","AI early warning and covenant monitoring is a monitoring system that flags borrowers whose credit is deteriorating across a loan book, for the relationship manager. It tracks covenant tests and borrower reporting, reads financials, filings and news, and combines them with payment and sector signals, then presents the evidence with a suggested next step.",[368],[21,435],"lending-and-credit",[437,438,379,439],"anomaly-detection","document-processing","summarization",[441,442,443,444],"Customers Bank","OakNorth Bank","PNC Financial Services","Sumitomo Mitsui Banking Corporation",{"slug":197,"title":446,"shortTitle":447,"definition":448,"status":9,"industries":449,"functions":450,"patterns":453,"audience":28,"autonomy":29,"adoptionStage":30,"evidenceCount":416,"publicEvidenceCount":416,"organizations":454,"bestGrade":240,"headline":299,"lastVerified":397,"indexable":232},"AI for tax compliance risk scoring and audit selection","Tax compliance risk scoring","Tax compliance risk scoring is AI that scores tax returns, taxpayers and transactions for the risk of error, underreporting or fraud, for a tax administration. It directs audit and compliance capacity to the highest risk cases, while an officer decides every compliance action and the selection itself is monitored for fairness.",[18],[21,451,452],"case-management","fraud-prevention",[23,437],[455,456,209,457,458,459],"Belastingdienst","Central Board of Indirect Taxes and Customs","Inland Revenue Authority of Singapore","Internal Revenue Service","South African Revenue Service",{"slug":198,"title":461,"shortTitle":462,"definition":463,"status":9,"industries":464,"functions":465,"patterns":467,"audience":28,"autonomy":29,"adoptionStage":30,"evidenceCount":383,"publicEvidenceCount":383,"organizations":468,"bestGrade":240,"headline":473,"lastVerified":397,"indexable":232},"AI for benefit fraud and error detection in social security","Benefit fraud and error detection","Benefit fraud and error detection uses risk models that help a social security or benefits agency decide which claims, payments and recipients to check for fraud or error. Caseworkers then verify the riskiest cases first, while every decision on entitlement stays with a person and the model is tested for fairness before and during use.",[18],[452,466,451],"citizen-services",[23,437],[469,247,470,471,472],"Centers for Medicare and Medicaid Services","Gemeente Rotterdam","U.S. Department of the Treasury, Bureau of the Fiscal Service","Uitvoeringsinstituut Werknemersverzekeringen (UWV)",{"kpi":474,"label":475,"unit":476,"n":349,"nUpTo":337,"kind":394,"value":477,"qualifier":270,"claimant":229,"organization":247,"vendorReported":210},"detection-rate-improvement","Detection improvement","multiplier",2.5,{"indexable":232,"reasons":479},[],[481,487,492,500,507,512,518,524,532,539,546,552,558,562,569,576,582,588,595,600,605,612,618,625,630,635,641,646,653,659,666,671,677,683,690,695],{"id":149,"label":482,"issuer":483,"region":157,"url":484,"description":485,"useCases":486,"indexable":232},"EU AI Act","European Union","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2024\u002F1689\u002Foj","Regulation (EU) 2024\u002F1689: risk based rules for AI systems, with obligations for high risk systems listed in Annex III and transparency duties under Article 50.",275,{"id":150,"label":488,"issuer":483,"region":157,"url":489,"description":490,"useCases":491,"indexable":232},"GDPR","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2016\u002F679\u002Foj","General Data Protection Regulation, including Article 22 on decisions based solely on automated processing.",246,{"id":493,"label":494,"issuer":495,"region":496,"url":497,"description":498,"useCases":499,"indexable":232},"iso-42001","ISO\u002FIEC 42001","ISO and IEC","global","https:\u002F\u002Fwww.iso.org\u002Fstandard\u002F81230.html","The international management system standard for AI.",143,{"id":501,"label":502,"issuer":503,"region":306,"url":504,"description":505,"useCases":506,"indexable":232},"nist-ai-rmf","NIST AI Risk Management Framework","NIST","https:\u002F\u002Fwww.nist.gov\u002Fitl\u002Fai-risk-management-framework","Voluntary US framework to map, measure, manage and govern AI risk, with a generative AI profile.",110,{"id":151,"label":508,"issuer":156,"region":157,"url":509,"description":510,"useCases":511,"indexable":232},"UK GDPR","https:\u002F\u002Fico.org.uk\u002Ffor-organisations\u002Fuk-gdpr-guidance-and-resources\u002F","The UK's version of the GDPR, including rules on solely automated decisions.",81,{"id":513,"label":514,"issuer":483,"region":157,"url":515,"description":516,"useCases":517,"indexable":232},"dora","DORA","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2022\u002F2554\u002Foj","Digital Operational Resilience Act for financial entities: ICT risk, incident reporting and third party risk, including AI providers.",72,{"id":519,"label":520,"issuer":162,"region":157,"url":521,"description":522,"useCases":523,"indexable":232},"uk-consumer-duty","FCA Consumer Duty","https:\u002F\u002Fwww.fca.org.uk\u002Ffirms\u002Fconsumer-duty","UK rules that require firms to deliver good outcomes for retail customers, including through automated channels.",51,{"id":525,"label":526,"issuer":527,"region":528,"url":529,"description":530,"useCases":531,"indexable":232},"mas-ai-risk-management","MAS AI risk management guidelines","Monetary Authority of Singapore","asia-pacific","https:\u002F\u002Fwww.mas.gov.sg\u002Fnews\u002Fmedia-releases\u002F2025\u002Fmas-guidelines-for-artificial-intelligence-risk-management","Singapore's supervisory expectations for AI risk management at financial institutions, building on the FEAT principles.",37,{"id":533,"label":534,"issuer":535,"region":528,"url":536,"description":537,"useCases":538,"indexable":232},"apra-cps-230","APRA CPS 230","Australian Prudential Regulation Authority","https:\u002F\u002Fwww.apra.gov.au\u002Foperational-risk-management","Australian operational risk standard covering critical operations and material service providers.",26,{"id":540,"label":541,"issuer":542,"region":306,"url":543,"description":544,"useCases":545,"indexable":232},"us-sr-11-7","SR 11-7 model risk management","Federal Reserve and OCC","https:\u002F\u002Fwww.federalreserve.gov\u002Fboarddocs\u002Fsrletters\u002F2011\u002Fsr1107.htm","US supervisory guidance on model risk management, applied by banks to AI and machine learning models.",23,{"id":547,"label":548,"issuer":483,"region":157,"url":549,"description":550,"useCases":551,"indexable":232},"nis2","NIS2 Directive","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Fdir\u002F2022\u002F2555\u002Foj","Directive (EU) 2022\u002F2555 on cybersecurity for essential and important entities, including telecom networks, energy and public administration.",22,{"id":553,"label":554,"issuer":555,"region":496,"url":556,"description":557,"useCases":551,"indexable":232},"pci-dss","PCI DSS","PCI Security Standards Council","https:\u002F\u002Fwww.pcisecuritystandards.org\u002F","Security standard for any system that stores, processes or transmits cardholder data.",{"id":152,"label":559,"issuer":167,"region":157,"url":168,"description":560,"useCases":561,"indexable":232},"UK Algorithmic Transparency Recording Standard","Mandatory transparency records for algorithmic tools used by UK central government.",18,{"id":563,"label":564,"issuer":565,"region":306,"url":566,"description":567,"useCases":568,"indexable":232},"hipaa","HIPAA","US Department of Health and Human Services","https:\u002F\u002Fwww.hhs.gov\u002Fhipaa\u002Findex.html","US rules for the privacy and security of protected health information.",17,{"id":570,"label":571,"issuer":572,"region":496,"url":573,"description":574,"useCases":575,"indexable":232},"fatf-recommendations","FATF Recommendations","Financial Action Task Force","https:\u002F\u002Fwww.fatf-gafi.org\u002Fen\u002Fpublications\u002FFatfrecommendations\u002FFatf-recommendations.html","Global standards for anti money laundering and counter terrorist financing that national rules implement.",15,{"id":577,"label":578,"issuer":483,"region":157,"url":579,"description":580,"useCases":581,"indexable":232},"eu-amlr","EU Anti Money Laundering Regulation","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2024\u002F1624\u002Foj","Regulation (EU) 2024\u002F1624: the single EU rulebook for customer due diligence, beneficial ownership and suspicious transaction reporting.",14,{"id":583,"label":584,"issuer":585,"region":306,"url":586,"description":587,"useCases":581,"indexable":232},"us-tcpa","Telephone Consumer Protection Act","Federal Communications Commission","https:\u002F\u002Fwww.fcc.gov\u002Fconsumers\u002Fguides\u002Fstop-unwanted-robocalls-and-texts","US consent rules for automated and prerecorded calls and texts; the FCC has confirmed AI generated voices count as artificial voices.",{"id":589,"label":590,"issuer":591,"region":306,"url":592,"description":593,"useCases":594,"indexable":232},"us-bsa","Bank Secrecy Act","FinCEN","https:\u002F\u002Fwww.fincen.gov\u002Fresources\u002Fstatutes-and-regulations\u002Fbank-secrecy-act","US anti money laundering law: customer due diligence, suspicious activity reports and record keeping.",13,{"id":596,"label":597,"issuer":483,"region":157,"url":598,"description":599,"useCases":594,"indexable":232},"eu-accessibility-act","European Accessibility Act","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Fdir\u002F2019\u002F882\u002Foj","Directive (EU) 2019\u002F882: accessibility requirements for banking services, ecommerce and other digital services, applicable since June 2025.",{"id":601,"label":602,"issuer":483,"region":157,"url":603,"description":604,"useCases":594,"indexable":232},"solvency-ii","Solvency II","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Fdir\u002F2009\u002F138\u002Foj","Directive 2009\u002F138\u002FEC: risk based capital, governance and model requirements for insurers.",{"id":606,"label":607,"issuer":608,"region":306,"url":609,"description":610,"useCases":611,"indexable":232},"us-ecoa-reg-b","ECOA and Regulation B","Consumer Financial Protection Bureau","https:\u002F\u002Fwww.consumerfinance.gov\u002Frules-policy\u002Fregulations\u002F1002\u002F9\u002F","US fair lending rules, including specific reasons in adverse action notices, which also apply when credit decisions use AI models.",12,{"id":613,"label":614,"issuer":615,"region":496,"url":616,"description":617,"useCases":611,"indexable":232},"telecom-consumer-rules","Telecom consumer protection rules","National telecom regulators","https:\u002F\u002Fwww.berec.europa.eu\u002F","National rules on telecom contracts, switching, billing disputes and marketing consent.",{"id":619,"label":620,"issuer":621,"region":157,"url":622,"description":623,"useCases":624,"indexable":232},"eba-loan-origination","EBA Guidelines on loan origination and monitoring","European Banking Authority","https:\u002F\u002Fwww.eba.europa.eu\u002Fregulation-and-policy\u002Fcredit-risk\u002Fguidelines-on-loan-origination-and-monitoring","Expectations for credit decisioning, including the use of automated models.",11,{"id":626,"label":627,"issuer":483,"region":157,"url":628,"description":629,"useCases":624,"indexable":232},"eecc","European Electronic Communications Code","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Fdir\u002F2018\u002F1972\u002Foj","Directive (EU) 2018\u002F1972: consumer protection, contract, switching and security rules for telecom operators.",{"id":631,"label":632,"issuer":483,"region":157,"url":633,"description":634,"useCases":624,"indexable":232},"eu-psd2","PSD2","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Fdir\u002F2015\u002F2366\u002Foj","Payment Services Directive 2: strong customer authentication, transaction risk analysis exemptions and open banking access.",{"id":636,"label":637,"issuer":527,"region":528,"url":638,"description":639,"useCases":640,"indexable":232},"mas-notice-626","MAS Notice 626","https:\u002F\u002Fwww.mas.gov.sg\u002Fregulation\u002Fnotices\u002Fnotice-626","Singapore's anti money laundering and counter terrorism financing requirements for banks.",10,{"id":642,"label":643,"issuer":483,"region":157,"url":644,"description":645,"useCases":640,"indexable":232},"mifid-ii","MiFID II","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Fdir\u002F2014\u002F65\u002Foj","Directive 2014\u002F65\u002FEU on markets in financial instruments: suitability and appropriateness of advice, record keeping and product governance.",{"id":647,"label":648,"issuer":649,"region":306,"url":650,"description":651,"useCases":652,"indexable":232},"us-fcra","Fair Credit Reporting Act","Federal Trade Commission","https:\u002F\u002Fwww.ftc.gov\u002Flegal-library\u002Fbrowse\u002Fstatutes\u002Ffair-credit-reporting-act","US rules on consumer reports, their accuracy and permissible use, relevant to credit scoring and screening.",8,{"id":654,"label":655,"issuer":483,"region":157,"url":656,"description":657,"useCases":658,"indexable":232},"eu-idd","Insurance Distribution Directive","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Fdir\u002F2016\u002F97\u002Foj","Directive (EU) 2016\u002F97: conduct rules for selling insurance, including demands and needs testing and advice.",7,{"id":660,"label":661,"issuer":662,"region":663,"url":664,"description":665,"useCases":383,"indexable":232},"cbuae-ai-guidance","CBUAE guidance on AI and ML","Central Bank of the UAE","middle-east","https:\u002F\u002Fwww.centralbank.ae\u002F","UAE central bank expectations for the enabling technologies, AI and machine learning used by licensed financial institutions.",{"id":667,"label":668,"issuer":483,"region":157,"url":669,"description":670,"useCases":384,"indexable":232},"eu-mortgage-credit-directive","EU Mortgage Credit Directive","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Fdir\u002F2014\u002F17\u002Foj","Directive 2014\u002F17\u002FEU: creditworthiness assessment, disclosure and advice rules for residential mortgage lending.",{"id":672,"label":673,"issuer":674,"region":157,"url":675,"description":676,"useCases":384,"indexable":232},"pra-ss1-23","PRA SS1\u002F23 model risk management","Prudential Regulation Authority","https:\u002F\u002Fwww.bankofengland.co.uk\u002Fprudential-regulation\u002Fpublication\u002F2023\u002Fmay\u002Fmodel-risk-management-principles-for-banks-ss","UK model risk management principles for banks, covering AI and machine learning models.",{"id":678,"label":679,"issuer":680,"region":157,"url":681,"description":682,"useCases":384,"indexable":232},"uk-psr-app-reimbursement","UK APP scam reimbursement rules","Payment Systems Regulator","https:\u002F\u002Fwww.psr.org.uk\u002Four-work\u002Fapp-scams\u002F","Mandatory reimbursement of authorised push payment scam victims by UK payment firms, which shifts scam losses onto banks.",{"id":684,"label":685,"issuer":686,"region":528,"url":687,"description":688,"useCases":689,"indexable":232},"au-scams-prevention-framework","Australian Scams Prevention Framework","Australian Treasury","https:\u002F\u002Ftreasury.gov.au\u002Fconsultation\u002Fc2024-573813","Economy wide obligations for banks, telcos and digital platforms to prevent, detect, disrupt and respond to scams.",3,{"id":691,"label":692,"issuer":483,"region":157,"url":693,"description":694,"useCases":689,"indexable":232},"eu-mar","EU Market Abuse Regulation","https:\u002F\u002Feur-lex.europa.eu\u002Feli\u002Freg\u002F2014\u002F596\u002Foj","Regulation (EU) 596\u002F2014: insider dealing and market manipulation, including the duty to detect and report suspicious orders and transactions.",{"id":696,"label":697,"issuer":698,"region":306,"url":699,"description":700,"useCases":689,"indexable":232},"nyc-local-law-144","NYC Local Law 144","New York City","https:\u002F\u002Fwww.nyc.gov\u002Fsite\u002Fdca\u002Fabout\u002Fautomated-employment-decision-tools.page","Bias audits and notices for automated employment decision tools used in hiring and promotion in New York City.",1791904360648]