For decades, the rhythm of tax practice has been organised around periods. Books are closed, figures are reconciled, returns are prepared, payments are made and questions are answered after submission. Monthly, quarterly and annual returns remain legally important, but the operating assumption behind that model - that tax compliance can be assembled mainly at the end of a period - is becoming increasingly fragile.

Business transactions are now created, approved and stored in digital systems. Fiscal devices, enterprise resource planning platforms, payroll applications, banking interfaces and tax portals can generate or receive information long before a practitioner opens the return. Tax administrations are also using richer data and more sophisticated analytics to identify inconsistencies, select cases and interact with taxpayers. The result is a decisive shift: compliance risk is increasingly visible at transaction level, not only at return level.

The profession should not respond by simply adding an artificial-intelligence tool to the old end-of-period routine. The stronger response is to redesign the routine itself. That means moving from periodic compliance to continuous compliance, supported by tax data assurance, governed technology and clear human accountability.

What continuous compliance really means

Continuous compliance does not mean submitting a tax return every day. Nor does it mean allowing a machine to make unreviewed tax decisions. It is the capability to keep tax obligations, transaction data, supporting evidence and control performance in a tax-ready state throughout the reporting cycle.

In a continuous model, tax-sensitive transactions are identified when they enter the business process. Required evidence is captured early. Data moves through controlled interfaces. Reconciliations and exception reports operate at an appropriate frequency. Unusual items are routed to a competent person, and decisions are documented before memories fade or staff change. Management receives visibility of exposures before they become filing crises.

The return still exists, but its role changes. Instead of being the place where compliance is first constructed, it becomes the formal output of a system that has been governed throughout the period. Deadlines remain important checkpoints; they are no longer the beginning of the compliance process.

Why the periodic model is reaching its limits

Periodic processes are naturally backward-looking. If a tax code is wrongly applied to thousands of transactions, a supplier's fiscal details are incomplete or a payroll configuration is incorrect, the error may accumulate for months before the return team discovers it. By then, correction may require system changes, document recovery, amended filings, interest, penalties or difficult explanations.

The return is also only as reliable as the information that feeds it. A technically correct tax interpretation cannot repair missing source data, weak access controls, undocumented spreadsheet changes or an interface that silently omits transactions. Tax risk therefore arises not only from misunderstanding legislation, but from the design and operation of the organisation's wider information environment.

Finally, tax authorities increasingly possess information from outside the return or from connected systems. A periodic process that asks only whether the taxpayer's return agrees to its own ledger may be too narrow. The more useful question is whether the transaction, fiscal record, accounting entry, supporting document, payment and reported amount form one coherent audit trail.

Zimbabwe's direction of travel

Zimbabwe is already moving toward a more connected compliance environment. ZIMRA's integration of the Tax and Revenue Management System (TaRMS) with the Fiscalisation Data Management System (FDMS) provides a clear example. ZIMRA's implementation notice states that input-tax data is made available to TaRMS through integration with FDMS, bringing fiscal invoice information closer to the VAT-return process.

This does not mean that every Zimbabwean tax obligation is currently assessed in real time, or that practitioners should assume that every advanced AI capability used internationally is already operating locally. It does, however, demonstrate the direction of travel: more structured data, stronger digital audit trails and greater ability to compare what was transacted, fiscalised, recorded and claimed.

For taxpayers, the practical consequence is that data quality can no longer be treated as an accounting issue to be cleaned up immediately before filing. For practitioners, it creates an opportunity to help clients strengthen processes before discrepancies reach the tax authority's systems.

Artificial intelligence changes both sides of compliance

The OECD reports that artificial intelligence is already used by more than 70 per cent of surveyed tax administrations, with common applications including the detection of evasion and fraud, risk assessment and virtual assistance. Its Tax Administration 3.0 work presents a broader vision in which tax processes become increasingly embedded in the systems taxpayers use to run their businesses.

For revenue authorities, AI can help prioritise cases, identify unusual patterns, match information and tailor services. For practitioners and taxpayers, similar methods can support transaction classification, reconciliation, duplicate detection, invoice validation, anomaly monitoring, deadline management and the organisation of evidence. Generative AI can also assist with summarising documents and producing first drafts of routine materials.

The important word is assist. AI output is not legislation, a binding interpretation or a defensible tax position merely because it is fluent. Models can rely on outdated material, invent citations, overlook facts, misread exceptions or reproduce errors in their training data. In tax, where one omitted condition can change the result, confidence of expression must never be mistaken for correctness.

Tax data assurance becomes the new professional backbone

Tax data assurance is the disciplined evaluation of whether the information used for tax is complete, accurate, valid, timely, consistent, secure and traceable. It is wider than checking the arithmetic on a return. It asks whether the data can be trusted from the point of origin to the final reported amount.

Completeness. The practitioner should establish whether all relevant transactions, entities, branches, employees, imports, payments and adjustments enter the tax process. Reconciliation must identify what is missing as well as what is present.

Accuracy and validity. Amounts, dates, tax codes, counterparty details and fiscal documents should be validated against authoritative sources and business rules. A field being populated is not proof that it is correct or legally sufficient.

Timeliness. Exceptions should be identified soon enough for the business to correct them before filing, payment or commercial decisions make remediation more difficult.

Consistency. The same transaction should be treated coherently across contracts, invoices, fiscal records, accounting systems, returns and payments. Differences must be understood and documented.

Traceability. A reported figure should be capable of being followed back through calculations and system interfaces to its source evidence, including the approvals and professional judgements applied along the way.

Security and control. Access to tax-sensitive systems and data should be role-based, changes should be authorised, and confidential taxpayer information should be protected when analytics or AI tools are used.

The future tax practitioner: six strengthened roles

Compliance architect

The practitioner helps design the path from transaction to return: identifying tax-sensitive events, assigning control owners, defining evidence requirements and ensuring that exceptions reach the right person. This moves professional input upstream, where errors are cheaper to prevent.

Tax data assurance specialist

The practitioner learns how data is created, transformed and reconciled. This does not require every tax accountant to become a software engineer. It does require enough systems and analytics literacy to ask the right questions, test interfaces and challenge incomplete evidence.

AI governance custodian

The practitioner establishes where AI may be used, which data may be entered, how outputs must be validated, who approves conclusions and how an audit trail is retained. The goal is not to block useful technology, but to make its use safe, explainable and professionally accountable.

Continuous risk monitor

Instead of waiting for a return, the practitioner reviews exception dashboards, unreconciled balances, invalid fiscal records, unusual effective rates and repeated control failures at a frequency matched to risk. Attention is directed to anomalies rather than dispersed across every transaction equally.

Interpreter and trusted explainer

As analytics become more sophisticated, clients need someone who can explain what a flag means, which facts matter, whether a rule has been correctly applied and what action is proportionate. Professional judgement becomes more valuable, not less, when automated outputs require context.

Dispute-readiness adviser

A continuous process preserves evidence and decision records before a query arises. The practitioner can therefore help the taxpayer respond faster, distinguish data mismatches from legal disagreements and present a coherent audit trail during verification, audit, objection or appeal.

Human judgement must remain visibly in control

An AI answer does not become a tax position until a competent professional has verified the law, tested the facts, considered contrary evidence and accepted responsibility for the conclusion.

Every tax practice using AI should have an acceptable-use framework. At minimum, it should address approved tools, data classification, client confidentiality, source verification, human review, documentation, retention, access, incident response and the circumstances in which AI use is prohibited. Staff should understand that copying confidential client information into an unapproved public tool can create a risk even when the resulting answer appears useful.

Practitioners must also guard against automation bias - the tendency to accept a system's answer because it appears objective. A risk score or anomaly flag should begin an inquiry, not end it. The taxpayer should be able to understand the data, rule and reasoning behind a material decision, particularly when the output may affect a filing, disclosure, dispute or client relationship.

A practical continuous-compliance operating cycle

Map obligations and data flows

Identify material taxes, filing obligations, transaction types, systems, interfaces, evidence sources and control owners. The map should show where tax-relevant information originates and every point at which it can be changed, omitted or delayed.

Validate at source

Build tax rules into onboarding, contracting, invoicing, procurement, payroll and fiscalisation processes where practical. Preventing an invalid transaction is more efficient than correcting thousands of records later.

Reconcile throughout the period

Use automated or controlled reconciliations between source systems, ledgers, fiscal records, tax portals, returns and payments. Set thresholds carefully so that material exceptions are visible without overwhelming reviewers with noise.

Analyse and route exceptions

Apply analytics or AI to identify patterns requiring attention, then route each issue to a named person. The workflow should distinguish data errors, control failures, uncertain legal positions and items that require management approval.

Decide, document and correct

A competent reviewer confirms the facts and law, records the conclusion, approves any correction and ensures that the root cause is addressed. Repeated exceptions should trigger process improvement, not repeated manual workarounds.

Report and retain evidence

Management receives a concise view of outstanding risks, control failures and remediation. Evidence of the transaction, calculation, review, decision, submission and payment is retained in an accessible and secure form.

A 90-day readiness plan for tax practices

Days 1-30: establish the baseline

Select one material tax area and map the end-to-end process. Inventory the data sources, spreadsheets, interfaces, portal access, review steps and recurring exceptions. Assess where the practice already uses AI or automation, including informal use by staff, and identify confidentiality or quality gaps.

Days 31-60: design controls and run a pilot

Define data-quality tests, reconciliations, exception thresholds, ownership and escalation. Approve a basic AI-use policy and train the engagement team. Pilot the process with one willing client or one internal compliance stream, using existing tools before investing in unnecessary technology.

Days 61-90: test, report and refine

Trace selected transactions from source to return, test whether controls operated and measure the volume and age of exceptions. Report findings to management, fix root causes and agree a recurring monitoring cycle. Document lessons before extending the model to other taxes or clients.

Continuous compliance must be proportionate

Continuous compliance is a control philosophy, not a requirement to purchase an expensive platform. A small practice or taxpayer may begin with a controlled calendar, secure document repository, standard data templates, protected spreadsheets, periodic reconciliations and a simple exception register. Larger organisations may justify integrated analytics, workflow systems and continuous control monitoring.

The frequency of monitoring should follow risk. A high-volume VAT or payroll process may require frequent checks, while a low-volume annual obligation may not. The principle is to identify and resolve material issues early enough to prevent accumulation, not to monitor every field merely because technology makes it possible.

New services and stronger professional relevance

This transition opens credible new areas of practice. Tax practitioners can offer tax-data health checks, continuous VAT or payroll assurance, fiscalisation and system-implementation reviews, tax-control frameworks, AI-governance policies, pre-audit data packs, anomaly-monitoring services and training for finance and operational teams.

These services strengthen rather than dilute the practitioner's technical role. Technology can identify a pattern, but it cannot by itself determine whether the pattern reflects an error, a lawful exception, a commercial reality or an uncertain interpretation. The practitioner supplies the legal knowledge, ethical responsibility, business context and explanation that turn information into sound action.

The return will remain - but it will no longer be where compliance begins

Periodic returns are not disappearing. They will remain important legal declarations and points of accountability. What is changing is their position in the compliance architecture. Increasingly, the quality of a return will be determined by controls that operated weeks or months earlier, inside systems and business processes that the return preparer may not directly control.

The future tax practitioner must therefore be technically authoritative, data-literate, technology-aware and visibly accountable. The profession's advantage will lie in combining what machines do well - scale, consistency and anomaly detection - with what professional judgement must do well: interpret law, test facts, challenge assumptions, protect confidentiality and explain consequences.

Practices that begin this transition now will be better positioned to prevent errors, respond to digital tax administration, defend client positions and offer higher-value advice. Those that remain centred only on periodic return preparation risk discovering that the most important compliance decisions have already been made upstream - by data, systems and controls they did not help to govern.