Modern tax debt management is increasingly driven by core revenue management systems (CRMS) that unify the taxpayer register, filing, payments, taxpayer accounting (ledger), and debt management into a single platform, typically exposed through a digital taxpayer account (portal/app). In Zimbabwe, ZIMRA’s Tax and Revenue Management System (TaRMS) is a prominent example: it replaces older domestic-tax systems and emphasizes automation, bank integration, and a “single account” taxpayer-ledger approach designed to reduce payment allocation errors and support automated debt detection, penalties/interest computation, reminders, and installment-plan applications.

At a policy level, the OECD’s tax-debt management work frames effective collection around four strategic principles: engagement before due date, maximizing collection after due date but before enforcement, effective enforcement, and realistic recovery/write-off. The same research highlights the shift toward e-services and contemporary digital channels (online payment plans, certified e-mail notifications, electronic auctions, and digital payment options) and the growing role of analytics, segmentation, and predictive models in choosing the right intervention for the right debtor at the right time.

Because the jurisdiction for your class is unspecified, this lesson treats TaRMS as a worked example (using official ZIMRA materials) while teaching a jurisdiction-neutral reference model for TaRMS-like systems. The lesson also foregrounds constraints: tax authorities handle highly sensitive data, so technology-driven debt management must align with tax secrecy/confidentiality rules, data protection/privacy laws, and governance requirements for automated decision-making/profiling (often including a right to human review for decisions with legal or similarly significant effects).

Assumptions made (explicit). - Jurisdiction is unspecified. The lesson therefore provides (a) an internationally common framework, (b) “common variants” across jurisdictions, and (c) a TaRMS (Zimbabwe) worked example because TaRMS has substantial public, primary documentation and is clearly a tax administration platform with debt-management features. - Access to vendor/implementation documentation is limited. Public materials describe TaRMS capabilities and taxpayer-facing workflows, but not full internal architecture, data models, or enforcement parameterization. Architecture and analytics sections therefore present a TaRMS-like reference architecture that you can adapt if you obtain internal manuals, system design documents, or procurement specs. - “Technology in debt management” is treated as the intersection of: (1) taxpayer accounts and communications channels, (2) automated workflow engines and business rules, (3) analytics/risk scoring, and (4) governance controls (privacy, security, due process).

How to localize for any jurisdiction (simple teacher method). - Replace the “Legal/Regulatory table” below with your country’s equivalents for (a) tax secrecy, (b) data protection, (c) e-service of notices, (d) automated decision-making safeguards, and (e) records management. Use the lesson’s system workflows and governance checklists unchanged.

The legal overlay for a tech-enabled debt system is usually “stacked”: tax administration law authorizes debt, penalties/interest, recovery powers, and communications; secrecy/confidentiality constrains disclosure; data protection/privacy constrains processing, sharing, retention, and automated decisions; and cybersecurity/identity rules shape authentication and secure integration.

Leading case law and regulatory guidance themes for automation and profiling

Because jurisdiction is unspecified, the lesson uses widely cited public-sector automation precedents and “regulator guidance” rather than a single country’s tax caselaw:

Robodebt Royal Commission (Australia) — cautionary precedent (cross-sector debt). The Australian government’s release summarizes the Commission’s finding that Robodebt was “neither fair nor legal” and notes a Federal Court finding of unlawfulness (late 2019). For a tax classroom, this is a high-impact case study on what can go wrong when automated debt processes are deployed without lawful basis, proper evidentiary logic, and robust human oversight.

SyRI (Netherlands) — risk-indication/profiling constraints (public fraud/benefits/taxes). The Netherlands’ judiciary communication on the SyRI case reports that the court found the SyRI legislation did not comply with Article 8 ECHR (privacy), highlighting proportionality and transparency concerns around government risk-profiling systems used for fraud detection in areas including taxes.

EDPB + ICO guidance — operationalizing safeguards. EU-level guidance (EDPB) and UK guidance (ICO) provide detailed interpretations of automated decision-making/profiling requirements, including what counts as “solely automated” decisioning and what rights/safeguards are expected (e.g., meaningful human review).

What TaRMS is and what it automates

ZIMRA states TaRMS is the new automation platform for domestic tax processes, replacing older systems and aiming to simplify processes for taxpayers through a Self-Service Portal (SSP).

Key TaRMS debt-management capabilities (officially described). TaRMS is described as enabling: automatic detection of unpaid liabilities, automated aging by set parameters, daily calculation of interest and penalties, automated reversals of interest/penalties on adjusted liabilities, automated reminders for payments and returns (before due date, on due date, and after due date), and online installment-plan applications.

Digital taxpayer account and ledger concept (“single account”). ZIMRA describes a “Taxpayer’s Single Account” record in TaRMS that tracks the taxpayer’s balance held in a ZIMRA single account at the taxpayer’s chosen bank; the system then handles payment allocations, assessments, and refunds based on submissions and integrated bank validation.

Reference architecture for TaRMS-like debt management systems

A TaRMS-like system typically includes these layers (this is a generalized architecture, anchored to observed TaRMS capabilities and OECD/IMF patterns):

Mermaid diagram: architecture and data flows for a TaRMS-like platform

This workflow connects (a) TaRMS debt automation, (b) OECD’s strategic principles, and (c) common operational controls (human review gates).

Debt creation and ledger posting. Liability is created by return submission, assessment, or adjustment; ledger is updated; TaRMS emphasizes automatic taxpayer ledger maintenance.

Pre-due engagement (prevent debt). OECD identifies early engagement before due date as a strategic principle; TaRMS describes automated reminders before due date.

Due date monitoring and “soft” collection. If unpaid, system detects unpaid liabilities and ages them; TaRMS explicitly describes automatic detection and aging of unpaid liabilities.

Automated charges. Interest and penalties are automatically calculated and posted (TaRMS: daily basis; penalties from due date).

Segmentation & channel selection. Analytics determine which message, which channel, and what escalation path is used; OECD notes analytics-supported segmentation and increasing automation for tailored engagement.

Self-service resolution. Encourage online payment plans and self-service tools; OECD and IRS examples show online payment plan creation and management.

Human review gates and enforcement. For high-impact actions (garnishment, liens, third-party collector handoff, insolvency referral), systems typically require human approval aligned with legal safeguards for automated decisions. OECD emphasizes enforcement tools as a principle; privacy guidance emphasizes safeguards for automated decisioning in impactful scenarios.

Procedural steps and timelines for automated triggers and human review

TaRMS confirms reminders occur prior to due date, on due date, and after due date and that liabilities are aged based on “set parameters,” but does not publish exact day offsets. Accordingly, use the following as a configurable teaching template (common patterns; adapt to local policy and system settings):

Mermaid diagram: automated debt workflow (from liability to enforcement)

Data sources for debt analytics and risk scoring

The IMF notes that analytics in tax administration are used for essential functions including monitoring filing and payment, analyzing arrears, segmenting taxpayers, and predicting behaviors. TaRMS sources (explicitly named) include integration with registries and banks for validation and payments workflows.

Risk scoring models: algorithms, inputs, thresholds, and bias mitigation

What tax administrations commonly do (OECD + IMF framing). - OECD’s maturity descriptors explicitly mention using data analytics, increased automation, and even predictive modeling and AI to select approaches for individual debtors as maturity increases. - OECD’s AI work notes that tax administrations use machine learning to score taxpayer behavior, detect anomalies, and prioritize cases for review—suggesting a “triage” function where automation routes cases but experts retain judgment for complex decisions.

A practical classroom risk-score design (teaching model, not legal advice). Define a score on 0–100 representing “collection risk” (higher = less likely to voluntarily pay / higher chance of default). A teachable approach is a two-stage model:

1) Propensity to pay / default model (predictive): Use logistic regression or gradient-boosted trees to estimate or . Typical inputs: - Debt size (current), debt age (days past due), count of overdue periods (recency/frequency) - Prior payment-plan history (ever, defaults) - Filing compliance (late filing frequency) - Sector/size proxies (turnover band if lawful/available) - Contactability (valid email/phone; portal activity) This aligns with the IMF’s description that analytics help monitor payment, analyze arrears, and predict behaviors.

2) Rules overlay (policy constraints): Convert predicted risk to action bands (thresholds) that reflect policy/legal constraints, for example: - S 0–30: automated reminders only; self-service plan auto-approval if requested - S 31–70: enhanced nudges + outbound phone queue; payment plan requires additional validation - S 71–100: caseworker review before any escalation; consider security, fraud checks, and hardship flags

This fits OECD’s principle of maximizing collection before enforcement and using tailored engagement based on segmentation.

Bias mitigation and “trustworthy AI” controls (minimum set). - Human-in-the-loop for high-impact decisions. OECD gives a concrete example outside tax debt: anomalies detected by AI are systematically verified by an official before action—a transferable governance pattern for enforcement decisions. - Avoid solely automated legal effects where rules prohibit it. GDPR Article 22 (and similar regimes) restricts decisions based solely on automated processing that have legal/similarly significant effects, requiring safeguards like human intervention in many scenarios. - Impact assessment before deployment. DPIA-type processes are widely recommended/required when using large-scale profiling, automated decision-making, or combining multiple data sources; the ICO lists these as DPIA triggers. - Fairness testing & drift monitoring. Evaluate error rates and action rates across legally relevant groups (where permitted) and across proxies (geography, sector) to detect disparate impacts; update models when behavior changes (economic cycles, policy changes). The SyRI case underscores that government risk systems face scrutiny on proportionality and privacy impacts. - Robust evidentiary logic. Robodebt is a powerful warning: large-scale automated debt activity that uses flawed logic can be found unlawful and cause large-scale harm, eroding trust.

Sample “features/costs/pros-cons” table for TaRMS-like systems

This table is illustrative for classroom comparison (use it to teach budgeting drivers when specific procurement costs are unavailable).

Learning objectives

Knowledge objectives (students can explain). Students will be able to explain: - What a TaRMS-like core revenue management system is and how it supports the debt lifecycle (ledger → detection/aging → reminders → payment plans → enforcement). - What a digital taxpayer account provides (view balances, notices, payment history, modify profile, request payment plans), using real-world examples from HMRC and IRS. - Why modern debt management relies on segmentation and analytics, and how predictive modeling/AI is used to tailor interventions. - The legal constraints that shape system design: (1) secrecy/confidentiality, (2) data protection, (3) automated decision-making safeguards, (4) security requirements.

Skills objectives (students can do). Students will be able to: - Map a taxpayer’s overdue-liability scenario into an automated workflow and identify where human review gates are required. - Design a basic risk scoring rubric (inputs, outputs, thresholds) and identify bias risks and mitigations. - Draft a compliant digital communication strategy (channel choice, content, cybersecurity warnings, escalation).

Suggested 75-minute class schedule

Context and framing (10 minutes). Prompt: “Why do tax debts accumulate, and why is digitalization changing collection?” Introduce OECD’s four strategic principles and connect them to platform capabilities (pre-due engagement + post-due escalation + enforcement + write-off).

TaRMS-like system walkthrough (15 minutes). Use TaRMS as the example: single account, automatic detection/aging, daily interest/penalties, reminders, online installment plan.

Hands-on workflow mapping (15 minutes). Students annotate the provided mermaid flowchart with where rules end and discretion begins (e.g., payment plan denial, enforcement triggers).

Analytics & governance mini-lecture (15 minutes). Explain segmentation and scoring using OECD/IMF framing; then introduce safety governance: human-in-the-loop, DPIA, and Article 22-style safeguards; use Robodebt as a cautionary case.

Group exercise: “Design a risk-based collection policy” (15 minutes). Groups design thresholds for three debtor segments and decide communications channels; then present tradeoffs.

Handouts (ready-to-print outlines)

Handout A: TaRMS-like debt management workflow checklist (1 page). - When does a liability post to the ledger? - What triggers the first reminder (pre-due vs post-due)? - When do interest/penalties begin? How are reversals handled on amended liabilities? - When is a payment plan offered / auto-approved? When must a caseworker decide? - Enforcement gate: “No high-impact action without human review + logged reasons.”

Handout B: Risk scoring design template (2 pages). - Define target outcome (pay within X days; plan default risk) - Features allowed vs prohibited (jurisdiction policy) - Thresholds mapped to actions - Bias and privacy safeguards checklist (DPIA, explainability, drift monitoring)

Handout C: Secure communications rules (1 page). - Prefer portal inbox for sensitive actions; avoid embedding payment/login links in SMS/email where possible - Standard phishing-report channels and taxpayer guidance (examples from HMRC/ATO/IRS)

Classroom hypotheticals with answers

Hypothetical 1: Automated reminders and debt aging (TaRMS features). A taxpayer files a return showing tax payable but does not pay by due date. Explain what a TaRMS-like system can do automatically and what it should not do without review.

Answer (model). Automatic steps include: detect unpaid liability, age it based on configured parameters, calculate and post interest daily and penalties from due date, and issue reminders per escalation procedures. Actions requiring review typically include high-impact enforcement (e.g., bank garnishment or legal proceedings) and any decision that denies relief or produces a significant effect, depending on jurisdictional automated-decision rules.

Hypothetical 2: Payment plan decisioning and Article 22-style safeguards. A risk engine automatically denies an installment plan to a taxpayer because the model predicts high default risk. The denial blocks access to a key relief mechanism.

Answer (governance). Treat this as potentially “significant” decisioning. In GDPR-like systems, a safeguard approach is: provide meaningful human review, allow the taxpayer to contest and present context, and document reasoning. A DPIA-style risk assessment is commonly required when deploying large-scale profiling/automation affecting access to a benefit/service; authorities should also test for bias and drift.

Hypothetical 3: Bias and proxy variables. The model uses “neighborhood” as a feature. It improves prediction accuracy but results in a higher enforcement referral rate for one socio-economic area.

Answer (mitigation). First, evaluate whether “neighborhood” is a proxy for protected status and whether disparate impact is occurring; second, consider feature restrictions or fairness constraints; third, ensure enforcement referrals require human validation and reasons. SyRI demonstrates courts may scrutinize government profiling for privacy/proportionality; robust transparency and safeguards matter.

Hypothetical 4: Automation failure scenario (cross-sector but transferable). A system generates large volumes of debt notices based on flawed calculation logic. What is the institutional risk?

Answer. The Robodebt Royal Commission demonstrates catastrophic risk: the scheme was described as neither fair nor legal, and the Federal Court found it unlawful. This illustrates why tax agencies must validate algorithms, maintain lawful bases, and implement oversight before scaling automated debt processes.

Mermaid decision tree: risk score → action with legal safeguards

MCQ (5).

1) In OECD’s strategic principles for successful tax debt management, which is listed as Principle 1? A. Immediate enforcement measures B. Engage with taxpayers before the due date C. Outsource all collections D. Stop using digital channels Correct: B.

2) TaRMS debt management (as described by ZIMRA) includes which automation? A. Automatic detection and aging of unpaid liabilities B. Automatic criminal prosecution C. Automatic court judgments D. Automatic debt forgiveness for all cases Correct: A.

3) Which is a feature explicitly listed in the IRS Online Account guidance? A. View digital notices and create payment plans B. File criminal complaints via the account C. Vote in elections D. Open a bank account at the IRS Correct: A.

4) What security control is specified in ZIMRA’s FDMS API documentation? A. HTTP only without encryption B. HTTPS only and (for most endpoints) mutual TLS client authentication C. No authentication required D. Passwords embedded in URLs Correct: B.

5) Which statement best reflects the Robodebt Royal Commission’s relevance to automated debt systems? A. Automation always increases fairness B. A large-scale automated debt scheme can be found unlawful and harm trust if logic and oversight are flawed C. Courts never review automated government systems D. Data protection laws do not apply to government processing Correct: B.

Short answer (3).

1) List five data sources a tax authority might use to score debt recovery risk and one governance risk for each. Expected elements: ledger/payment history, filing history, digital engagement, registry/bank signals, enforcement outcomes; governance concerns: secrecy/privacy, accuracy, bias, security, lawful basis.

2) Explain why “human review gates” matter in automated collections. Expected elements: prevents solely automated legal effects without safeguards; supports contestability; reduces error propagation; improves trust.

3) Describe TaRMS “single account” concept in 3–5 lines. Expected elements: taxpayer chooses one bank; payments go to ZIMRA single account; TaRMS records taxpayer single-account balance; system allocates to liabilities and supports assessments/refunds.

Essay prompts (2).

1) “Risk scoring improves efficiency but threatens fairness and legality.” Discuss, proposing a governance framework for tax debt analytics. A-grade coverage: OECD analytics/segmentation maturity; EDPB/ICO safeguards; DPIA; human review; explainability; audit logging; lessons from Robodebt and SyRI.

2) Design a TaRMS-like automated collections playbook for a mid-sized tax authority, specifying triggers, channels, and review points. A-grade coverage: pre-due engagement; post-due reminders; payment plan pathways; enforcement gating; secure communications and anti-phishing measures.