Did OpenAI Just Become a Living Example of the Fusion Governance Framework?

The Real-Time Tax Briefing — Issue 1

For decades, enterprise finance and tax functions operated with an implicit temporal cushion for post-facto reconstruction. A transaction occurred on Day 1; it was recorded, routed, adjusted, reconciled and reported over the subsequent 30 to 90 days. Where accounting required correction or a tax treatment required further review, downstream human intervention — a month-end close, a manual journal entry or a tax provision true-up — gave the organisation time to repair the record before external reporting.

That cushion is shrinking fast.

Tax administrations are increasingly adopting e-invoicing, continuous transaction controls and other forms of digital transaction reporting. Regulatory visibility is consequently moving closer to the underlying transaction. At the same time, enterprise finance is moving in the same direction internally, shortening the path from source data to management understanding and decision-making. The OECD's Tax Administration 3.0 vision captures a similar movement towards tax processes becoming more deeply embedded in the systems businesses already use.

When OpenAI CFO Sarah Friar published her reflections on building an AI-native finance function, her account offered a striking real-world illustration of what the Fusion Governance Framework describes. OpenAI approaches the challenge from finance and technology. Real-time tax approaches it from regulatory visibility and transaction governance. Both journeys are moving towards a common operating reality: the traditional gap between an event, its interpretation, its control and its visibility is becoming much smaller. In that sense, OpenAI's finance function reads like a living example of the Fusion Governance Framework, viewed from the enterprise side rather than the tax side.

What OpenAI's Finance Build Actually Demonstrates

Friar's account of OpenAI's finance transformation highlights four operating principles with direct relevance for modern tax governance: a zero-day close ambition that delivers continuous financial visibility with a dramatically shorter reconstruction cycle; continuous reconciliation of purchase orders, actuals, accruals and spending plans; AI-enabled workflows grounded in approved sources, with controlled changes to approved financial baselines; and technology accelerating execution while people retain responsibility for the outcome.

These principles extend beyond finance efficiency. They echo a tax environment increasingly built around structured transaction data and faster regulatory visibility. For the enterprise, this brings an additional requirement: data, evidence, interpretation and accountability all need to move at a similar speed.

The Tax Velocity Gap™: When the Authority Sees Faster Than You Reconcile

The first implication is what the Fusion Governance Framework calls the Tax Velocity Gap™: the shrinking interval between an economic transaction and its structured visibility to a tax authority. Friar outlines OpenAI's pursuit of a zero-day close and continuous forecasting, with the ambition of giving leadership a real-time, traceable view of the company's financial position as business conditions change. As authority-facing visibility accelerates, an organisation faces growing risk when its own understanding of the same transaction continues to depend on a month-end reconstruction cycle.

The issue is larger than reporting speed; it is an information asymmetry. External systems may receive structured transaction data while internal teams are still reconciling, interpreting or correcting the underlying activity. The strategic response is to ensure internal understanding keeps pace with external transaction visibility.

Controls Move Upstream: The Continuous Controls Environment™

That response starts at the point of transaction. OpenAI is working towards connecting approved spending plans, general-ledger actuals, purchase orders, accruals and transaction details into a continuously reconciled foundation. Friar's description shifts the role of the close towards reviewing exceptions and applying judgment to an increasingly connected financial view.

This closely reflects the logic of a Continuous Controls Environment™ (CCE). Traditional control models placed significant reliance on review and correction after transactions had already been recorded. As transaction processing and external visibility accelerate, controls need to move closer to source. Tax determination, master-data quality, classification and supporting evidence increasingly become part of the transaction process itself. Control therefore shifts from retrospective repair towards upstream execution, with the mandate to embed tax determination, master-data validation and key compliance controls as close to transaction origin as practical.

Tax-as-Code: Which Interpretation Is Authorised to Become Executable?

Once controls move upstream, the knowledge that drives them must also become explicit. Friar describes specific workflows grounded in approved materials, alongside governed forecasting baselines where changes require finance authorisation. The underlying principle is important: machine-supported execution works best when the knowledge, assumptions and boundaries governing it are explicit.

This mirrors Tax-as-Code: the translation of governed tax rules and approved interpretations into explicit, version-controlled system logic. Domain expertise gains greater leverage when approved interpretations can travel through systems consistently. A position that once lived in an email, spreadsheet or individual memory can increasingly become part of the logic that determines how thousands of transactions are processed. That creates a governance question with lasting significance: which interpretation has been authorised to become executable, who owns it, what evidence supports it and how will future changes be governed? The strategic mandate is to translate governed tax rules and approved interpretations into version-controlled system logic wherever execution can be standardised.

Executable Is Not Defensible: Evidence Architecture

Even with strong upstream controls and executable knowledge, a critical question remains: can the enterprise explain and defend what it has done? Friar describes interactive finance tools that connect forecasts, explanations and scenarios back to underlying account-level evidence and business activity. As quantitative reconciliation becomes increasingly automated and continuous, qualitative reconciliation takes on greater importance. The enterprise still needs to establish whether a classification, business rationale, legal interpretation or tax treatment is supported by sufficient evidence.

This distinction matters in real-time tax, particularly as authorities gain earlier and more structured access to transaction data. Systems can become highly efficient at matching numbers while the more difficult questions concern meaning: whether the transaction has been classified correctly, whether the supporting facts justify the treatment applied and whether the enterprise can reconstruct that reasoning later. Tax-as-Code makes a position executable; Tax Defensibility & Evidence Architecture™ makes it defensible. The strategic mandate is to connect material tax positions to the underlying commercial facts, documents and evidence required to explain and defend them.

Two Governance Questions Raised by Machine-Supported Execution

Human accountability in automated workflows

In the IR-GPT example, technology produces a first draft rapidly, while the investor relations team reviews the response, adds context and retains responsibility for the final communication. The same principle becomes critical in tax. As automated and agentic workflows take on larger portions of execution, responsibility can become distributed across systems, models, data owners, process owners and professional teams. A strong governance model keeps accountability visible throughout that chain. Systems may execute. Human and organisational accountability remains explicit. Autonomous workflows therefore need clear decision rights around interpretation, approval, escalation and defence. The greater the scope of machine execution, the more important it becomes to define who owns the meaning and the outcome.

Outcome assurance over technical validation

Friar encourages CFOs to measure the dependable work technology completes and the quality of the resulting decisions, instead of relying on adoption metrics such as licence counts or token usage. The same discipline applies to real-time tax. Technical transmission represents one layer of compliance. Substantive assurance asks whether the transaction carries the correct legal, tax and economic meaning. A compliance dashboard may show "network green" because the file transmitted successfully and passed schema validation. The underlying treatment may still carry "semantic red" because a product, customer, tax code or legal interpretation was wrong. This is the assurance challenge in a highly automated environment: technical success and substantive correctness need to be assessed separately.

The External Dimension: What Leaves the Enterprise

Friar's model focuses primarily on creating faster and more trustworthy intelligence inside the enterprise. Real-time tax introduces an external dimension. Machine-produced classifications and transaction data can leave the enterprise, become visible to authorities and persist as part of an evidentiary record. That creates additional governance questions around semantic correctness, external representation and replayability.

An internal forecast can be revised before management acts on it. An externally transmitted tax record may already sit within an authority-facing environment before the enterprise completes its own retrospective review. That difference makes upstream correctness, evidence and accountability materially more important in real-time tax. It also changes the way enterprises need to think about transaction truth. The accounting view, tax view, commercial reality and externally transmitted representation increasingly need to remain connected throughout the life of the transaction.

Governed at Real-Time Speed

Sarah Friar's account shows that the move towards real-time operations has progressed well beyond theory. Finance teams are actively redesigning workflows around continuous information, connected evidence, faster analysis and clearer human ownership. For tax leaders, the lesson extends beyond the adoption of new technology. The same architecture that enables faster financial intelligence also exposes the weaknesses of tax processes that still depend heavily on downstream reconstruction.

Using technology to accelerate a fragmented process may improve cycle time. Greater value comes from redesigning the operating model upstream so that data quality, tax logic, evidence, controls and accountability travel with the transaction from the beginning. This changes the economics of governance. Real-time systems extend the window in which an enterprise can influence the future while compressing the window available to repair the past. Tax teams therefore need to move beyond period-end review and participate directly in the design of data, systems, controls and decision logic.

In that sense, OpenAI's finance function can be read as a living example of the Fusion Governance Framework from the enterprise side. The next step for tax is to ensure the same architecture governs what leaves the enterprise, not just what informs management. Speed and control now need to move together. The ultimate competitive and regulatory advantage belongs to the organisation that is governed at real-time speed.

The Fusion Governance Framework Architecture

This article applies four of the eleven pillars of the Fusion Governance Framework to the intersection of AI-native finance and real-time tax: the Tax Velocity Gap™, the Continuous Controls Environment™ (CCE), Tax-as-Code, and Tax Defensibility & Evidence Architecture™. These sit within a broader architecture that also addresses inbound tax assurance, Cognitive Responsibility Diffusion™, outcome assurance and the emerging role of Fusion Professionals™.

A full description of the framework architecture, including additional pillars and implementation guidance, is available at nitin-agarwal.com/frameworks.

From Principle to Practice

At Contiqa Systems, this is the operating problem we help businesses solve. We work with organisations to move tax governance upstream by connecting data, tax logic, ERP design, controls, evidence and accountability across the transaction lifecycle.

Our work spans e-invoicing readiness, transaction and data diagnostics, tax determination, ERP and technology architecture, control design, evidence governance and post-go-live assurance. The objective is to help businesses build an operating environment in which transaction data is reliable at source, tax logic is governed, controls operate at the right point in the process, and externally visible outcomes can be explained and defended.

As finance, tax and technology become increasingly connected, the challenge is moving beyond individual compliance projects towards an architecture that can operate continuously. Contiqa helps businesses build that governance layer — so that speed, automation and regulatory visibility can scale together.

This is Issue 1 of The Real-Time Tax Briefing. Future issues will land in subscribers' inboxes before they are announced anywhere else — subscribe below to get Issue 2 first.