TABLE OF CONTENTS
Free Learning Resource
Artificial Intelligence
AI & Tax A Guide for Tax Leaders
I. Introduction: The Transformation of Tax Through AI
Over the past two decades, the tax function has undergone a significant transformation. What was once a predominantly manual, compliance-driven activity has evolved into a digitally enabled and increasingly data-intensive function. The introduction of enterprise resource planning (ERP) systems, tax engines, and e-filing platforms has improved efficiency, standardization, and control. Yet, despite these advancements, the core operating model of tax has remained largely unchanged: human-driven processes supported by technology.
​
This paradigm is now shifting.
​
Artificial Intelligence (AI) represents more than an incremental technological improvement. It marks a structural inflection point in the evolution of tax. Unlike traditional rule-based systems, AI introduces adaptive, learning-based capabilities that enable tax processes to move beyond automation towards intelligent—and ultimately autonomous—execution.
​
This shift is driven by several structural developments that are fundamentally reshaping the tax environment.
Digital and Real-Time Regulatory Landscape
First, the regulatory landscape has become increasingly digital and real-time. Tax authorities are moving away from retrospective reporting towards continuous transaction-level visibility. Real-time reporting regimes, e-invoicing systems, and continuous transaction controls are becoming standard across jurisdictions. These developments reflect a broader transformation described in initiatives such as the OECD's vision for digital tax administration and EU reforms including VAT in the Digital Age (ViDA). As a result, tax functions must operate with a level of speed, granularity, and accuracy that traditional models were not designed to support.
Expanding Scale and Nature of Tax Data
Second, the scale and nature of tax-relevant data have expanded dramatically. Tax is no longer confined to structured financial data. It now encompasses a wide range of data sources, including contracts, legal interpretations, transfer pricing documentation, and cross-border transactional flows. Much of this data is unstructured and fragmented across systems. AI—particularly large language models—enables organizations to extract, interpret, and connect these data sources in ways that were previously not feasible.
Pressure to Deliver More with Less
Third, tax functions are facing increasing pressure to deliver more with limited resources. Organizations expect tax departments not only to ensure compliance, but also to provide forward-looking insights, support business decisions, and proactively manage risk. This creates a structural imbalance between complexity and capacity that cannot be addressed through incremental improvements alone.
AI offers a pathway to address this imbalance. However, its true significance lies not only in efficiency, but in transformation.
​
AI enables tax functions to shift from reactive, compliance-oriented units to proactive, intelligence-driven capabilities embedded within business operations. In this emerging model, tax is no longer a downstream function processing historical data, but an integral part of real-time decision-making.
​
Ultimately, the question is no longer whether AI will transform tax. That transformation is already underway. The critical question is how organizations will adapt their operating models to harness its full potential.
II. What Is AI in Tax?
Artificial Intelligence in tax refers to the application of machine-based systems capable of performing tasks that traditionally required human judgment within tax processes. These tasks include analysing transactional data, interpreting regulations, generating documentation, and supporting decision-making. Increasingly, AI systems are also capable of executing elements of tax workflows, moving beyond assistive roles into operational functions.
​
However, AI in tax is not simply about introducing new tools. It reflects a broader shift in how tax functions are designed and how they operate.
​
Traditional tax technology relies on deterministic logic and predefined rules. Systems such as tax engines and ERP integrations are designed to process structured data within fixed parameters. While highly effective for standardized processes, these systems are inherently limited in their ability to deal with ambiguity, interpret complex regulations, or adapt to continuous regulatory change—challenges that are central to tax.
Pattern recognition & learning
Context interpretation
Structured & unstructured data
Adaptive insights & recommendations
AI introduces a fundamentally different paradigm. Rather than relying solely on static rules, AI systems learn from data, identify patterns, and interpret context. This enables them to handle both structured and unstructured information, including contracts, legal texts, and jurisdiction-specific guidance. As a result, tax processes can evolve from rigid, rule-based workflows into adaptive systems that continuously improve over time.
​
This shift also changes the nature of outputs. Traditional systems generate predefined results based on input parameters. AI systems, by contrast, can generate insights, recommendations, and draft decisions. For example, instead of merely calculating a tax outcome, an AI-enabled system can assess alternative treatments, highlight uncertainties, and suggest the most appropriate approach based on available data.
​
This evolution can be understood as a progression of capabilities: from descriptive, to diagnostic, to predictive, to prescriptive, to autonomous.
​
At the descriptive level, AI supports data extraction and classification. At the diagnostic level, it identifies anomalies and inconsistencies. At the predictive level, it anticipates risks and outcomes. At the prescriptive level, it recommends actions and optimizes decisions. At the autonomous level, AI systems are capable of executing end-to-end tax processes with limited human intervention. While most organizations are still positioned in the earlier stages, the shift towards higher levels of capability is accelerating.
​
Importantly, AI operates as an embedded intelligence layer within the broader tax technology architecture. This architecture includes data pipelines (ETL processes), ERP and tax systems, process frameworks, and governance structures. AI connects and enhances these components, enabling tax functions to operate more efficiently, respond more quickly to regulatory changes, and manage complexity at scale.
​
In this sense, AI does not replace existing tax technology—it transforms how it is used.
​III. Why AI Is Becoming Critical for Tax Functions
The growing importance of AI in tax is driven by structural changes that are reshaping the tax environment. Regulatory pressure, data complexity, and increasing expectations are converging in ways that make traditional approaches increasingly unsustainable.
​
Tax authorities worldwide are transitioning towards real-time reporting models supported by digital infrastructures such as e-invoicing and continuous transaction controls. These developments are not isolated—they are part of a broader shift towards fully digital tax administration environments, as outlined in OECD frameworks and EU regulatory initiatives. Compliance is no longer periodic; it is continuous, data-driven, and increasingly automated.
​
At the same time, tax authorities are becoming more technologically advanced. The use of data analytics, automated risk assessment, and cross-system data matching is now standard practice. This significantly increases the likelihood of detecting inconsistencies and potential non-compliance. In this environment, reactive compliance models are no longer sufficient. Tax functions must be able to monitor, analyse, and respond to risks in near real time.
​
The nature of tax-relevant data further reinforces this challenge. Tax functions must now process a wide range of structured and unstructured data sources, including financial data, legal documents, and transfer pricing analyses. This data is often fragmented across systems and jurisdictions, making it difficult to manage using conventional tools. AI enables tax functions to integrate and analyse these data sources more effectively, transforming them into actionable insights.
​
Compounding these challenges is a growing imbalance between complexity and capacity. Tax teams are expected to manage increasing regulatory and operational complexity without proportional increases in resources. This creates a structural gap that cannot be addressed through traditional process improvements alone.
​
At the same time, expectations of the tax function are evolving. Tax is increasingly seen as a strategic capability within organizations. Businesses expect tax teams to provide forward-looking insights, support decision-making, and contribute to value creation. This requires capabilities that extend beyond compliance and reporting.
​
AI provides the foundation for this transition. By automating repetitive tasks, enhancing data processing, and supporting predictive and prescriptive analysis, AI enables tax functions to scale their operations and expand their role within the organization.
​
Taken together, these developments make one conclusion clear: AI is no longer optional. It is becoming a foundational component of a tax function capable of operating effectively in a digital, real-time, and data-driven environment.
​IV. The Evolution of AI in Tax: From Hype to Autonomy
AI in tax is evolving from experimental applications towards integrated and increasingly autonomous systems. Organizations are currently positioned at different stages of this evolution, depending on their technological maturity and readiness.
​
This development can be understood as a progression across five stages: Hype, GenAI+, AGI, Superintelligence, and Singularity. While this staged model provides a practical framework for understanding AI maturity in tax, it is consistent with broader perspectives on how AI transforms professional services. As emphasized by Richard Susskind, the impact of AI lies in its ability to reshape how expertise is created, distributed, and embedded within systems.
Stage 1: Hype
In the Hype Stage, organizations experiment with isolated AI tools without fundamentally changing their processes. AI is explored as a novelty, with limited integration into existing tax workflows.
Stage 2: GenAI+
The GenAI+ Stage marks the integration of AI into specific workflows, particularly in areas such as research, data processing, and documentation. AI becomes a practical tool within defined use cases.
Stage 3: AGI
A more fundamental shift occurs at the AGI Stage, where AI systems begin to support complex reasoning and decision-making across jurisdictions. This enables more integrated, end-to-end tax processes.
Stage 4: Superintelligence
The Superintelligence Stage further expands these capabilities, enabling advanced scenario modelling, real-time optimization, and continuous risk monitoring.
Stage 5: Singularity
The final stage—Singularity—envisions fully autonomous tax systems capable of executing end-to-end processes with minimal human intervention. While still emerging, elements of this stage are already visible in agent-based systems and advanced automation.
As discussed in How to Think About AI, the long-term trajectory of AI involves embedding expertise into systems rather than individuals.
V. From Automation to Agentic Tax Workflows
As AI continues to evolve, the tax function is moving beyond traditional automation towards a new paradigm: agentic workflows. While automation focuses on executing predefined tasks more efficiently, agentic systems introduce a fundamentally different approach—one in which AI-driven entities can perform, coordinate, and optimize tasks within a broader workflow.
​
This distinction is critical. Automation improves existing processes; agentic workflows begin to redefine them.
​
In a traditional tax environment, workflows are typically linear and human-driven. Tasks are assigned, executed, reviewed, and completed in a sequential manner. Even when supported by technology, the structure of the process remains largely unchanged. AI, in its earlier stages, fits into this model as a tool that enhances specific steps—such as data extraction or document drafting—without altering the overall workflow logic.
​
Agentic workflows, by contrast, operate on a more dynamic and distributed model. Instead of relying on static task sequences, workflows are orchestrated through multiple AI agents, each responsible for specific functions. These agents can analyse data, make decisions within defined parameters, interact with other agents, and adapt their actions based on context. The result is a system that is not only automated, but also responsive and self-adjusting.
Rather than assigning a tax professional to manually review transactions, determine the correct treatment, and prepare documentation, an agentic system could distribute these tasks across specialized AI agents.
​
In the context of tax, this can be illustrated through a simple example. Rather than assigning a tax professional to manually review transactions, determine the correct treatment, and prepare documentation, an agentic system could distribute these tasks across specialized AI agents. One agent may analyse transactional data, another may interpret relevant regulations, while a third may generate documentation or flag uncertainties. These outputs are then consolidated and presented to a human professional for validation and oversight.
​
This model introduces a new form of collaboration between humans and technology. Instead of performing tasks directly, tax professionals increasingly supervise, validate, and guide AI-driven processes. AI systems generate outputs at scale, while humans focus on ensuring quality, managing exceptions, and exercising judgment in complex or ambiguous cases. This shift allows tax functions to operate more efficiently while maintaining control and accountability.
​
A key feature of agentic workflows is their ability to dynamically allocate tasks. Rather than relying on fixed assignments, tasks can be distributed based on capacity, complexity, and priority. This approach can be compared to platform-based models seen in other industries, where tasks are continuously reassigned to optimize performance. In a tax context, this could mean that a set of tax-related tasks is automatically allocated across a combination of AI agents and human professionals, with the system continuously adjusting based on workload and outcomes.
​
Implications of Agentic Workflows
​
-
Tax functions can scale more effectively, as tasks are no longer constrained by linear workflows or fixed resource allocations
-
Consistency improves, as AI agents apply standardized logic across large volumes of data
-
Responsiveness is enhanced, allowing tax functions to adapt more quickly to regulatory changes and emerging risks
​
However, the introduction of agentic workflows also raises important considerations. Governance becomes more complex, as decision-making is distributed across multiple systems rather than centralized in human actors. Ensuring transparency, traceability, and accountability is therefore essential. Organizations must establish clear frameworks for oversight, define the boundaries within which AI agents can operate, and implement mechanisms to monitor and validate outcomes.
​
Despite these challenges, the direction of travel is clear. Tax functions are moving from process-driven models towards system-driven ecosystems, where workflows are no longer rigid sequences, but adaptive networks of human and machine collaboration.
​
In this emerging model, the question is no longer how to automate individual tasks, but how to design and orchestrate entire tax workflows in which AI and human expertise are seamlessly integrated. This marks a fundamental shift—from automation as a tool, to agency as a capability.
​
Governance of Agentic Workflows
​
The increasing autonomy of agentic tax workflows also introduces new governance challenges. As decision-making becomes distributed across systems and AI agents, traditional control mechanisms are no longer sufficient. This requires a more structured governance approach, which can be conceptualized as a layered model.
1
Governance Principles
At the foundation, organizations must establish clear governance principles, including data integrity, transparency, and human oversight, ensuring that AI systems operate within defined ethical and regulatory boundaries.
2
Operational Orchestration
At the operational level, governance shifts towards the orchestration of tools and workflows, defining how AI systems interact, how decisions are validated, and where control points are embedded within the process.
3
Capability and Accountability
At the highest level, effective governance depends on capability and accountability—ensuring that individuals interacting with AI systems possess the required expertise and that responsibilities for oversight are clearly assigned.
In this context, governance is not merely about control; it becomes a mechanism to enable trust, scalability, and consistency in increasingly autonomous tax environments.
VI. Conclusion: The Rise of Autonomous Tax Functions
The transformation of tax through AI is no longer a forward-looking concept—it is already underway. What began as experimentation with isolated tools is evolving into a broader shift in how tax functions are structured, operated, and integrated within organizations.
​
As AI becomes more deeply embedded into tax processes, the function itself is changing in nature. Tasks that were once manual and time-intensive are increasingly handled by intelligent systems. Workflows are becoming more dynamic, adaptive, and interconnected. Most importantly, decision-making is gradually moving closer to real-time, supported by data-driven insights rather than retrospective analysis.
​
This evolution does not eliminate the role of tax professionals. On the contrary, it elevates it. As AI takes over execution-heavy activities, human expertise becomes more focused on interpretation, judgment, and oversight. The value of the tax function shifts from processing information to shaping decisions—ensuring that outcomes are not only technically correct, but also aligned with regulatory expectations and business objectives.
​
At the same time, the emergence of agentic workflows signals a deeper transformation. Tax is no longer defined by linear processes, but by systems that can coordinate, adapt, and operate with a degree of autonomy. This marks the beginning of a transition towards what can be described as autonomous tax functions—environments in which data, technology, and intelligence are seamlessly integrated into a continuous operating model.
​
For organizations, this shift presents both an opportunity and a challenge. Those that successfully adopt AI will be able to manage complexity at scale, reduce risk, and position tax as a strategic capability within the business. Those that delay may find it increasingly difficult to keep pace with regulatory developments and the growing sophistication of tax authorities.
​
Ultimately, the question is no longer whether AI will transform tax. That transformation is already in motion. The critical question is how organizations choose to respond.
​
Whether they adapt their operating models to harness AI effectively, or remain constrained by approaches that are no longer sufficient in a digital, data-driven environment.
​
The future of tax will not be defined by tools alone, but by the ability to embed intelligence into every layer of the function. In this future, tax is not simply compliant—it is connected, responsive, and increasingly autonomous.
Resource Hub — Continue Your Deep Dive
Your next steps in mastering AI in tax — from structured courses to cutting-edge research.