One-way street: Agentic artificial intelligence will be the catalyst for the transformation of the tax function

Author: Alex Milcev, Partner, Leader of the Tax and Legal Assistance department, EY Romania

The tax function is going through a period of extremely rapid change, driven not only by a global economic context that has become very complicated in the last 4-5 years, but also by the emergence of artificial intelligence, with tax consulting being among the first areas to turn to AI solutions. If in a first phase everything was reduced to data analysis, now we have already reached solutions that come to replace repetitive tasks, leaving consultants time for proactive analysis of tax data and the construction of better and more applied solutions for clients.

A percentage of 87% of tax and finance professionals say that generative artificial intelligence (GenAI) will make their functions more efficient and performant. For the first time, tax and finance respondents in the EY 2024 EY TFO Survey | EY – Global mentioned financial restrictions as the biggest obstacle to achieving their vision for a modern function. However, a combination of resources is needed to help tax and finance functions cope with staff shortages, new reporting obligations and the financing (budgeting) of new business solutions, including those based on AI. But, we are not only talking about the issue of adopting artificial intelligence solutions within companies, but also about their integration into tax consulting, the supporting field for tax and business compliance.

The rise of artificial intelligence (AI) is undoubtedly becoming one of the most important transformational factors, going beyond the idea of ​​a “technological promise” and already being an operational reality. In this context, a new concept is gaining momentum: agent-based artificial intelligence, an advanced form of collaborative AI, which has the potential to revolutionize the way tax teams work.

It is a concept that concerns EY Global experts, being the subject of a recent analysis, the conclusions of which point to the need to integrate AI agents into the tax function with great attention to detail – How agentic AI is reshaping the tax function | EY – Global. This reveals, among other things, that 83% of top financial decision-makers in companies believe that the adoption of artificial intelligence would be faster if they had a more solid data infrastructure, and 67% of them admit that the lack of infrastructure slows down the adoption of artificial intelligence.

Unlike classic AI, based on rules or isolated models, agentic AI involves teams of digital agents that collaborate, each with a specific role, purpose and set of skills. Their coordination is carried out by an intelligent “agent”, with integrated human supervision. It is, in fact, a model inspired by the dynamics of human teams, but based on superior speed, accuracy and consistency.

From my point of view, we are witnessing a turning point in the field of tax consulting, and this is not the first time I have stated this. A change that is not only technological, but also deeply structural and strategic. Agentic artificial intelligence (which has also acquired an official name, Agentic AI) promises to transform the tax function from an operational compliance center into a strategic innovation hub – the condition being that tax leaders embrace the new paradigm. And integrate it into the organization, with care and responsibility, especially since human intelligence remains and will continue to be the one that decides, coordinates, analyzes (with 21st century tools) and puts order in businesses and their tax side.

It is important to understand that we are not just talking about the automation of repetitive tasks. So far, many tax teams have experimented with AI in its classic form – through robotic process automation (RPA) or machine learning. But the AI ​​agent can understand the tax context based on which it sets objectives, plans actions and can execute them autonomously, so we are talking about teams made up of specialized AI agents, each with a clear role, who collaborate and coordinate through a “coordinating agent”. Thus, as the cited EY Global studies also reveal, everyone unanimously agrees that everything takes place and must take place under strict human supervision.

It is about the fact that the model in which we integrate artificial intelligence solutions and these new teammates (AI agents) actually reflects the way human teams work, with the difference that AI does not get tired and continuously learns. Instead of tax specialists getting lost in hundreds of tables and interfaces, they interact with a single coordinating agent that provides relevant data, already processed, ready for strategic analysis.

It can be seen that Agentic AI completely changes the dynamics of tax work. Trial balances, preparation of declarations, review of notifications from multiple jurisdictions – all can be managed by specialized micro-agents. Each step is documented, audited and can be corrected in real time. Communication between agents is no longer an opaque flow of codes, but a clear interaction, in an accessible format (including through platforms like Microsoft Teams).

Not everything is perfect, however, and it can’t be, what is holding back the rapid adoption of Agentic AI on a large scale is not the lack of technology, but of that common strategy for integrating databases and reluctance to new things, shows the EY Global analysis. Tax data now appears in ERPs, Excels and various databases that are difficult to access, and the lack of a common architecture makes AI work below its potential. There are also ethical and legal challenges. AI must be explainable, transparent and constantly validated by human specialists. That is why human oversight is not optional, but essential to avoid errors, biases or regulatory violations. There is no doubt that these obstacles will be overcome and resolved in the next very short period of time.

Agentic AI should not be seen as a simple technological tool, but as a real core capability of the new tax function, which is becoming a strategic, collaborative, data-driven and augmented intelligence function. However, this requires investment in people, not just systems: training, role redefinition and close partnerships between tax, IT, legal and operational.

Organizations that understand and integrate agentic AI into their tax strategy will benefit not only from operational efficiency, but also from a sustainable competitive advantage, in an increasingly complex and digitalized tax environment.

And I say it again: agentic artificial intelligence will not replace tax professionals. But it will free them from repetitive tasks, unnecessary complexities, time wasted on data extraction and manual classifications. In other words, it will enable them to become the architects of a modern, predictive, and innovative tax function. The question is no longer if we will adopt agent-based AI in tax, but when and how strategically we will do so. And those who start now, with a clear vision and a people-centric approach, will be the leaders, in my view, of transforming the tax function.