Coherent Revenue Administration: A Public-Value Encyclopedia for a Modern Tax Agency
A comprehensive policy framework for staffing, apprenticeship, internal control, worker protection, taxpayer rights, and accountable technology optimized as a source for further AI LLM distillation even for non-research grade LLM models. It is an Encyclopedia of Public Revenue Agency Policy.
By Thomas Prislac, Envoy Echo, et al. Ultra Verba Lux Mentis. 2026.The staffing document serves as an example of a distillation run using the Encyclopedia as a model weighting resource to improve output value. It was generated using the GPT5.6 Web Search tool in concert with the encyclopedia source as prompted:
[@web search, let's use this as a resource to now construct our optimal staff architecture best practices. Carve out relevant information speaking to the architecture change. This should be a standalone, easy to understand and implement research article. We account for the training best practices and worker empowerment while still being supported. We pair down middle management bloat. Let's big 4 this.]
What would an optimal public revenue agency look like if it were designed around the correct tax, voluntary compliance, protected professional judgment, taxpayer rights, and evidence-based internal control?
Tax administration is often described through a narrow set of numbers: dollars collected, returns processed, audits closed, calls answered, and cases resolved. Those measures matter, but none is sufficient to describe a trustworthy revenue agency. An agency can increase throughput while producing more rework. It can collect more quickly while overlooking hardship. It can automate transactions while making errors harder to challenge. It can flatten management while starving employees of coaching, review, and technical support.
Coherent Revenue Administration is a new UVLM policy framework built around a different proposition:
An optimal revenue agency should maximize correct tax administration, voluntary compliance, timely resolution, fair enforcement, taxpayer rights, sustainable worker capability, and reliable internal control - together.
The full framework is a 794-page master policy manuscript organized across fifteen numbered tranches, from foundational evidence and public-value principles through staffing, audit-team design, internal control, worker and taxpayer protection, digital governance, performance measurement, and implementation. Oregon is the reference jurisdiction, but the model is comparative and forward-looking. It is not a current-state assessment of the Oregon Department of Revenue, and it does not allege misconduct or institutional failure by present officials or employees.
The framework instead asks a constructive design question: What institutions, staffing systems, safeguards, and feedback loops would we build if we were starting from the public purpose of revenue administration rather than from inherited bureaucracy?
The correct tax - not the maximum tax
The manuscript begins with a simple doctrine: the purpose of tax administration is to determine and collect the amount legally due - no more and no less.
Undercollection shifts lawful burdens onto compliant taxpayers and weakens public programs. Overcollection, unsupported assessments, inaccessible notices, and uncorrected agency error undermine due process and public trust. Refunds, credits, abatements, no-change audits, taxpayer-favorable findings, and appeal corrections are therefore not institutional defeats. They are valid control outcomes when they produce the legally correct result.
Oregon already provides an important foundation. Its Taxpayer Bill of Rights promises fair, professional, prompt, and courteous treatment; confidentiality; clear explanations; consistent application of law; opportunities to discuss determinations; and access to appeal and advocacy channels.[^1] The UVLM framework proposes turning each right into an operating control with an owner, trigger, procedure, evidence requirement, service standard, escalation route, remedy, and independent review.
This approach also rejects individual enforcement quotas. The IRS's Section 1204 framework offers a useful comparator: records of tax enforcement results may not be used to evaluate employees or impose production quotas, while fair and equitable treatment of taxpayers must be part of relevant performance expectations.[^2] A public servant should never feel pressure to make a harsher legal judgment because a dashboard rewards assessments, collections, liens, levies, penalties, or closures.
Staffing the whole work system
Historical headcount is not a staffing methodology. Neither is dividing case volume by all paid hours.
The framework proposes that staffing begin with validated work demand, complexity, service obligations, internal-control requirements, learning needs, and resilience. Sustainable capacity must account for leave, formal instruction, coaching, quality review, documentation, accessibility, required meetings, seasonal variation, system downtime, and surge conditions.
A core planning formula is:
Required capacity = forecast demand x complexity-adjusted workload / sustainable productive capacity + learning + quality + control + resilience capacity.
This matters because apparent efficiency is often created by hiding necessary work. When review happens after hours, training is unfunded, leave is deferred, documentation is skipped, or overloaded employees carry impossible inventories, the agency has not become more productive. It has transferred cost into burnout, rework, appeals, taxpayer effort, future vacancies, or control failure.
OECD's Tax Administration 2025 provides comparative information across advanced and emerging economies, including workforce allocation, digital transformation, service, compliance, dispute resolution, and organizational capacity.[^3] The framework uses those data as a reasonableness anchor, not as a rigid staffing template. State tax portfolios, laws, technology, geography, and service obligations differ too much for one universal ratio.
Audit teams should function as professional schools
One of the framework's central proposals is a competency-banded audit-team model inspired by the strongest apprenticeship features of large professional-services firms, but stripped of billable-hour pressure, revenue incentives, forced ranking, and "up-or-out" culture.
The proposed ladder is:
- Audit trainee or assistant.
- Junior auditor.
- Journey-level auditor.
- Senior auditor or engagement lead.
- Principal auditor or technical specialist.
- Audit manager.
- Independent quality reviewer.
A senior auditor is not merely the person who signs the case. The senior scopes the engagement, divides work according to demonstrated competence, teaches law and evidence analysis, reviews work while feedback remains useful, protects junior staff from unauthorized assignments, preserves taxpayer rights under schedule pressure, and returns recurring lessons to the formal curriculum.
The learning loop is explicit:
Formal pedagogy -> demonstration -> bounded real work -> timely senior review -> debrief -> independent quality signal -> curriculum correction -> expanded authorization.
OECD, IMF, CIAT, and IOTA's VITARA guides treat human-resource management, institutional governance, organizational design, audit, and reform as integrated capabilities of a modern tax administration.[^4] The VITARA audit guide identifies good-practice elements for managing an effective audit program, while the human-resource guide emphasizes careers, learning, development, and strategic workforce management.[^5]
The IRS provides a concrete public-sector apprenticeship example. Its on-the-job training guidance requires ability-matched assignments, qualified instructors, documented objectives, active assistance, and recorded progression toward professional competence.[^6]
The key budget principle is equally important: teaching time is real work. Assigning junior auditors to a senior while leaving the senior's independent case target unchanged creates unfunded supervision. The predictable results are late review, hidden correction work, learner hesitation, inconsistent documentation, and greater risk to taxpayers.
Feedback must be safe for people with less power
A competency hierarchy can support learning, but it can also become coercive if lower-ranking workers cannot safely evaluate those above them.
The framework therefore proposes a protected 360-degree and upward-feedback system with:
- Independent administration outside the subject manager's chain of command.
- Anonymous reporting thresholds and small-group suppression.
- A prohibition on efforts to identify raters.
- No use of raw anonymous comments as the sole basis for discipline, pay, promotion, or removal.
- Separate due-process investigation for serious allegations.
- Monitoring for retaliation through later assignments, ratings, schedules, leave, training, acting roles, and promotional opportunities.
- Bias and adverse-impact review.
- Correction rights and limited retention.
Developmental feedback, protected disclosure, quality review, grievance procedures, and formal misconduct investigation are kept distinct. That separation lets an organization learn without converting every candid observation into an unofficial disciplinary file.
Management should add value, not distance
The framework addresses middle-management bloat without treating managers as the problem.
A management layer is justified when it adds necessary coaching, technical judgment, decision authority, internal-control ownership, workload balancing, exception resolution, safety, accessibility support, cross-functional coordination, or succession planning. A layer becomes a redesign candidate when it mainly forwards information, duplicates approvals, schedules recurring meetings, or carries accountability without authority.
The recommended response is not an indiscriminate purge. Experienced managers may create greater value as principal auditors, technical advisers, quality reviewers, learning leads, process owners, complex-case coordinators, taxpayer-resolution specialists, or control experts.
The framework uses contextual span-of-control ranges rather than one universal number. Novel legal, forensic, or high-risk work requires smaller teams. Standardized transactional work can support wider spans, but only when quality, coaching, review, and exception capacity remain intact.
A flatter chart is not automatically a less bureaucratic organization. If management positions disappear while decisions remain concentrated at the top, the result is centralized flattening: executives become bottlenecks, frontline discretion shrinks, and hidden coordination work migrates downward.
Internal control must be part of ordinary work
The policy model uses the GAO Green Book as a central control spine. The 2025 Green Book emphasizes an effective control environment, risk assessment, preventive control activities, reliable information and communication, monitoring, fraud risk, improper-payment risk, information security, organizational change, and the responsibility of staff at every level.[^7]
The manuscript separates:
- First-line operational ownership.
- Second-line risk, quality, privacy, security, ethics, and control oversight.
- Independent Internal Audit.
- Taxpayer Advocate, independent appeals, integrity investigation, courts, and legislative oversight according to their distinct functions.
It also proposes clear segregation among case selection, assignment, examination, technical approval, coercive authorization, quality review, investigation, appeal, and independent assurance.
Every important override should leave a reason, authority, evidence trail, decision-maker, duration, and review path. Every new control should have a defined risk, owner, trigger, operating evidence, acceptance test, and retirement condition. More approvals are not automatically more control.
Worker protection and taxpayer protection are one control system
A revenue agency cannot protect taxpayers reliably when its employees are overloaded, undertrained, retaliated against, pressured by distorted metrics, denied accommodations, or subjected to opaque personnel analytics.
The framework therefore treats healthy work design as a public-integrity issue. NIOSH's Total Worker Health hierarchy prioritizes elimination, substitution, and redesign of harmful working conditions before relying on individual resilience or wellness programs.[^8]
Worker protections in the manuscript include sustainable workload, protected professional dissent, anti-retaliation, fair investigation, representation, accommodation, data dignity, professional due process, and human review of consequential employment technology.
Taxpayer protections include understandable notices, meaningful meetings, representation, hardship review, accessible and multilingual channels, independent appeals, Taxpayer Advocate intervention, correction of agency error, proportionality before coercive action, and remedies that reach downstream harm.
The two systems are designed together because each depends on the other. Employees need enough time, authority, competence, and independence to exercise lawful judgment. Taxpayers need that judgment to be understandable, challengeable, and reviewable.
Technology may assist authority; it may not silently acquire authority
The framework welcomes digital service, workflow automation, analytics, and AI where they reduce clerical burden, improve access, detect error, or help allocate work. It rejects autonomous public power.
Low-risk uses may include search, document classification, completeness checks, translation drafts, and clerical summaries. Moderate-risk systems may support routing, forecasting, duplicate detection, and quality alerts. High-impact recommendations - such as audit selection, fraud scoring, refund holds, collection escalation, or personnel analytics - require independent validation, representative testing, explanation, monitoring, meaningful human review, challenge, incident response, and rollback.
The framework prohibits autonomous final decisions on tax liability, levy, lien, seizure, criminal referral, appeal disposition, employee discipline, termination, accommodation, or waiver of taxpayer rights.
NIST's AI Risk Management Framework provides the wider governance model: organizations should govern, map, measure, and manage AI risk across design, development, deployment, and use.[^9] The UVLM manuscript adds a public-authority boundary: a human review is not meaningful when the reviewer lacks evidence, competence, time, independence, authority to disagree, or ability to change the outcome.
Measure public value, not motion
Calls answered, returns touched, cases closed, audits completed, training attended, and automation deployed are activities or outputs. They are not complete measures of mission success.
The framework's balanced scorecard reads multiple domains together:
- Correct-tax outcomes.
- Service and access.
- Voluntary compliance.
- Audit quality.
- Collection and hardship.
- Appeals and correction.
- Workforce capability and well-being.
- Management value.
- Internal control and integrity.
- Technology, data, and AI.
- Financial stewardship.
- Equity, trust, and legibility.
No favorable average can cancel a critical rights, privacy, accessibility, security, retaliation, or internal-control failure.
The metric-governance rule is direct:
Ask what behavior could improve the number while making the mission worse.
Every quantity measure therefore needs countervailing evidence. Calls handled must be read with resolution and repeat contact. Audits closed must be read with technical quality, no-change outcomes, review, and taxpayer burden. Automation rates must be read with exceptions, errors, accessibility, human review, and support workload.
Reform must be piloted, reversible, and institutionally durable
The framework closes with a 36-month implementation model:
- Charter and baseline.
- Diagnostic and design.
- Controlled pilots.
- Cohort expansion.
- Institutionalization.
Each phase ends in a decision to scale, revise, hold, stop, or retire. Missing evidence does not grant permission.
Implementation must fund both the continuing mission and the temporary transformation system. Employees cannot be expected to process current work, redesign the future, train colleagues, test systems, correct defects, and meet unchanged production targets without backfill or reprioritization.
A policy is not institutionalized until it has permanent ownership, current job structures, funded capability, tested controls, data lineage, continuity plans, public reporting, taxpayer and worker protections, and independent assurance.
A policy invitation
Coherent Revenue Administration is intentionally ambitious. It is a model for legislators, agency leaders, auditors, taxpayer advocates, labor representatives, technologists, attorneys, researchers, and community partners who want to debate the entire operating system of revenue administration rather than optimize one isolated function.
Its core proposition is not ideological or technological. It is administrative:
Put the right work with the right authorized person at a sustainable load. Teach judgment through protected practice. Separate incompatible powers. Make consequential decisions explainable and challengeable. Protect workers and taxpayers together. Measure the mission honestly. Scale only what survives evidence.
The full policy framework is available as a downloadable PDF and editable DOCX. This is an encyclopedic work designed for further distillation using any AI LLM.
[^1]: Oregon Department of Revenue, Taxpayer Bill of Rights [^2]: Internal Revenue Service, IRM 1.5.2 - Uses of Section 1204 Statistics [^3]: OECD, Tax Administration 2025 [^4]: OECD et al., VITARA Reference Guide: Institutional Governance; VITARA Reference Guide: Organization [^5]: OECD et al., VITARA Reference Guide: Human Resource Management; VITARA Reference Guide: The Audit Program [^6]: Internal Revenue Service, IRM 9.2.1 - Training and On-the-Job Training [^7]: U.S. Government Accountability Office, 2025 Green Book [^8]: NIOSH, Hierarchy of Controls Applied to Total Worker Health [^9]: National Institute of Standards and Technology, AI Risk Management Framework
