The skills paradox: Why tax leaders must surrender what they prize most

The most dangerous liability in a tax function today is not its technology debt or its regulatory exposure. It is people's attachment to the very skills that built their careers.

The language of AI in general, and agent architecture in particular, reveals a critical talent and transformation challenge. In today’s AI systems, a “Skill” is defined as a repeatable capability an agent can execute, such as flagging transfer pricing anomalies, identifying incentives, or drafting technical memoranda. This is not a metaphor. It is a structural identity. For decades, the tax profession has largely determined career progression in precisely the same terms: the accumulation of discrete, repeatable capabilities that marked you as competent, then senior, then an expert.

The human skill and the machine Skill are the same unit of productive capacity, expressed in different forms. And that identity is now colliding with a transformation need that tax leaders can no longer defer. PwC research shows more than 80% of tax leaders expect AI to reshape tax planning and strategy within three years. At the same time, 95% of tax professionals report skills gaps in their functions, even as those same skills are increasingly replicable in systems. Similarly, The World Economic Forum's Future of Jobs 2025 report projects that 39% of existing skill sets will be disrupted within five years, with analytical and tech skills being both the most displaced and most in-demand. The numbers tell a contradictory story. Most leaders see AI reshaping the function. And yet the skills gap keeps widening—not despite AI adoption, but perhaps, in part, because of it. The skills being replaced aren't being replaced by new ones fast enough, either via agents or the talent capabilities.

The leaders who recognize this tension—and act on it—will redesign their organizations. Those who resist will find their functions redesigned for them. The opportunity is the willingness to treat transformation as a design challenge, not a threat to manage.  

When an organization deploys an AI agent within its tax function, it builds that agent the same way it once trained a junior associate: task by task, skill by skill. The agent is given a scope—a provision calculation, a regulatory research query, a compliance task—and trained to execute it with consistency, speed, and auditable precision. Layer enough “Skills” together, and you begin to replicate portions of workflows that were once performed entirely by tax professionals.

The human development side has always operated on the same architecture. You learned the code. You mastered the calculations. You earned your seat at the table by accumulating skills through repetition—and advanced as your expertise, and how you applied it in context, exceeded what others could offer. Hiring criteria, training programs, performance reviews, and promotion decisions all indexed to this logic—evaluating not just what you knew, but how you operated.

Once a skill can live equally well in a person or a system, the leadership question shifts from “what skills do we have?” to “where should those skills live?” This is an allocation problem that has a clear resolution principle: comparative advantage. This is not an argument for elimination but for precision in allocation—ensuring each skill sits where it delivers the greatest advantage, human or machine. Done well, this unlocks human capacity for higher-order work—judgment, context, and consequence—where we create the most differentiated value.

Tax functions are capacity constrained by compliance volume, resource constraints, and a deluge of data, limiting their ability to deliver strategic value. Continuing to allocate human capital to repeatable, system-executable work carries a clear opportunity cost and that cost is increasingly visible. For tax, that means acknowledging an uncomfortable truth: many of the skills that once defined career trajectories are now agent-class work. Not inferior work. Not unimportant work. Tasks that machines can handle just as well as people —which means human capital is freed for work that benefits from being done by a human: the work organizations have always needed but could never create the space for or get enough of.

The implication is direct: let go of work you can do to concentrate on work only you should do. What that higher-order work looks like—and what it requires—is what makes this transformation worth pursuing.  

If this were merely a technology deployment problem, most tax functions would have solved it already. PwC is observing that across industries, AI investment is widespread, but progress on achievement of targeted outcomes is not. The real obstacle in this transformation is organizational identity.

Tax professionals built their careers by owning specific skills. Those skills became a large part of their value proposition—the reason they were hired, the architecture of their departments. Asking a subject matter expert, a specialist partner or a Head of Tax to encode those skills into an agent is asking them to confront their own relevance. And identity questions are where transformation efforts go to die.

Institutional structures reinforce the problem. Organizations are built around skills ownership, including hierarchies, metrics, and training. When that work shifts to machines, the structure breaks. This helps explain why so many AI initiatives stall at the pilot stage. The technology works, but the organization, and its people, cannot absorb it.

What is underway is the commoditization of specific forms of human capital. It has happened before: in legal discovery, in financial trading, in manufacturing. In each case, the professionals who moved up the value chain thrived. Those who defended their incumbency did not. There is no reason to believe tax will be the exception, but it is also early enough to be on the right side of what is needed. Herminia Ibarra’s work on Working Identity finds that identity change happens through action – you don’t think your way into a new professional identity, you act your way into it.  

None of this implies a diminished profession. Quite the opposite. When routine human skills migrate to agents, what remains is the work that has always mattered most but was perpetually crowded out by volume: judgment, synthesis, risk positioning, strategic counsel. PwC’s Global Reframing Tax Survey reinforces this shift: tax is increasingly expected to play a central role in business strategy and reinvention, not just compliance.

The critical distinction is that AI primarily transforms tasks and how work is performed, not the people who perform it. It does not dictate a single workforce outcome. Leadership choices about where to invest, what work to automate, and how to redeploy talent ultimately shape the future model. This is not a story about elimination. It is a story about elevation.

The tax professional of the future is an orchestrator. They design which Skills agents execute, interpret outputs with contextual judgment about a business's risk appetite and regulatory posture, translate tax implications into language that moves C-suite decisions, and govern the system to ensure efficiency does not come at the expense of defensibility. As an example, an employee reviews an AI-generated transfer pricing analysis, applies judgment about a client's specific regulatory posture and gets guidance from a tax leader on prior or other context considerations. They collectively reframe the risk for a CFO conversation and sign off on the workflow design that generated it. They are not doing less. They are doing what only they can do. They operate alongside an integrated team of engineers, data scientists, and technologists. The hierarchy of specialists gives way to a multidisciplinary unit.

The workforce strategy that supports this shift starts with a straightforward principle: AI should expand human potential. That means clear pathways to future-ready skills like AI fluency, advanced analytics, critical thinking, and advisory capabilities, and careers built around insight and problem-solving, not data assembly and reconciliation. The talent pipeline must evolve in parallel. Entry-level training today is increasingly agent-class work. Organizations must rethink what "entry-level" means, recruiting from disciplines that develop reasoning and ambiguity tolerance, and investing in those people with the infrastructure to build future-ready careers. This is not a story in which entry-level roles disappear. It is a story in which they transform — toward reasoning, judgment, and ambiguity tolerance from day one, rather than data assembly and reconciliation.

Professional services firms are the proving ground for this transition. They are training the next generation in AI-integrated environments, rotating them across capabilities and industries, and embedding technologists directly within tax engagements to work alongside tax professionals, and creating the new fluencies that the profession will need. The talent being forged inside these firms will carry these new fluencies into every organization they eventually service or join, reshaping the profession from the inside out. For the mid-career professional, the path runs through deliberate upskilling – for themselves and their teams. This requires leaning in to build AI fluency and data interpretation while refining and coaching more junior team members on critical thinking, advisory communication and relationship development, which requires that organizations treat that investment as a priority, not a footnote.  

The tax function will be remade. The skills that once anchored careers are migrating to systems that execute them faster, more consistently, and at scale. The question before every tax leader is not whether to adapt but what to build and agentify to best complement the talent development opportunities now available.

For an individual in the field: Audit your skills portfolio with ruthless clarity. Map every recurring task and ask: is this a capability that should remain human, or one that should be encoded? What remains—the technical capability beyond the page, judgment, orchestration, strategic counsel—is the foundation of a tax function that is not just more efficient but genuinely more valuable. For those who lead tax functions: the same audit applies at the organizational level. Which capabilities belong in your people, and which belong in your systems? The answer to that question is your transformation roadmap.

The distinction between a human skill and a machine Skill is dissolving. The leaders who treat that dissolution as a design opportunity will build something their predecessors could not have imagined and will define what tax leadership looks like for the next generation.

Transformation will not wait. The only question is whether you will be its architect, or its artifact.