PwC Romania Annual Tax Conference 2026: Technology and AI are reinventing the tax function and bringing substantial time savings, but the need for governance is greater than ever
The fiscal function in Romania is going through an accelerated transformation under the simultaneous pressure of the digitalization of reporting obligations and the rapid evolution of artificial intelligence. This was the central theme of the panel dedicated to tax technology within the PwC Romania Annual Tax Conference.
Watch the recording of the discussion here:
New regulations such as SAF-T, e-Invoice and e-Transport have made localization of ERP systems mandatory, not optional. In practice, however, many companies run into limitations of group systems or systems that cannot be easily configured for local requirements. Accounting and tax teams remain under constant pressure between increasing volumes of data, increasingly complex reporting and increasingly limited resources.
In this context, the localization or migration process of the ERP system should no longer be seen only from a technical perspective or alignment with group requirements, but should consider local reporting requirements. Accounting and tax teams must be involved in this process to ensure that the new ERP system covers as many of these requirements as possible.
Where full localization is not feasible, an alternative is gaining ground: the managed services model. An external provider fully takes over the client's accounting or tax function, providing not only human resources, but also access to technology and automation without additional investment – from intelligent invoice processing to algorithm-generated bookkeeping proposals.
But there is also a third way: in situations where localization only partially covers local requirements, and managed services are not a viable option, companies can turn to the integration of technology into internal flows. Specifically, this involves identifying repetitive, time-consuming accounting and tax processes and integrating technology solutions to take them over.
On the artificial intelligence front, the discussion highlighted the shift from predictive AI to autonomous agents capable of executing complex end-to-end processes such as automatic document reconciliation and interpretation of legislative changes. The uses of AI are multiple and even countless, we just need to identify those situations where AI can help us and explore how we get the best results. The main barrier to adoption, however, is not the technology, but the way people interact with it: the quality of prompts and personalized training make the difference between spectacular results and the perception that it “doesn't work”.
However, given that AI has become an “invisible colleague” in many organizations, risks are also materializing: biases in automated decisions, credible but false hallucinations, loss of confidential data through the uncontrolled use of publicly available tools. The AI Act, the first European legislation dedicated to artificial intelligence, provides stringent requirements for high-risk systems and the penalties are severe. Responsible adoption of AI is not optional, but strategic.
The main statements of the panel:
Silvian Milea, PwC Romania Partner: Digital transformation is no longer a one-off project but a continuous process
We live in a world where the pressure for efficiency, speed and adaptability is at an all-time high. We see a continuous acceleration of the digitization of tax reporting obligations, we see a very rapid development of technology and artificial intelligence. All of this has a significant impact on the tax function and the way we work. We need to be nimble, make quick and informed decisions while managing ever-increasing volumes of data, processes and expectations, both internal and external. Digital transformation is no longer a one-off project, it's no longer what we do this year – it's actually an ongoing process.
Until now you knew us primarily for tax consulting, but we have also transformed. We have crossed the boundaries of traditional services, developed new categories of products and services that integrate technology, automation and intelligent management of tax data. For us it is important to be with our customers where they need it most. Today we are talking about some of the essential pillars of digital transformation – about accounting management systems, which have long been no longer just a place where we entered data, but tools that help us work faster and more efficiently, about process optimization through technology and about artificial intelligence.
I give an example from our experience: when registering invoices, we have a module that takes the invoice – it can come via e-Invoice, scanned in PDF, by email, in any format -, sends it to an AI that looks at it, recognizes which supplier it is from, what the amounts are, automatically sends it on an approval flow to the customer, to different approvers depending on the type of invoice, and once approved, automatically creates the supplier as well. And then an AI comes, looks at how the invoices were recorded in the past and what the law says, and proposes the accounting record as well. Obviously, a human has to look at it, but the process of accounts payable, of recording invoices, is simply faster, more flexible and more robust from the point of view of control.
Anca Macovei, Director of PwC Romania: When we have a question related to something in accounting policy, we no longer search in hundreds of pages but ask the AI assistant
Major projects such as SAF-T, e-Invoice and e-Transport have propelled the digitalization of fiscal obligations, and we realize that we can no longer do it manually. Who could do the SAF-T manually? We are talking about a lot of data, in a lot of reports, mostly from the same sources. Moreover, we are also talking about a transmission of this data in standardized formats, in real time or in a very short period after the monthly close. We'd all like to have a SAF-T declaration at the click of a button, wouldn't we? How could we fulfill this desire? By having a system that complies with all local requirements, by correct accounting records from the start and without doing manual processing outside the system. Localization is clearly no longer optional, and accounting and tax teams must be involved in the process
We have a lot of legislative changes, quite frequent, and it would help us and our clients to know as quickly as possible what changed, when it changed, what the impact is, what we need to report. We have developed AI agents on specific categories of legislation to help us interact with legislation much faster and know exactly what to report and when. It works and works very well. Another example: we have a lot of internal procedures and policies, a lot of documents. We have developed and implemented agents that help us interact with internal policies. Think how good it is when we have a question about something in accounting policy, and instead of searching through dozens or hundreds of pages, we ask our internal policy AI assistant how best to solve the problem.
If we look at our discussion of technology and AI today, I think the key message is clear: technology and AI are no longer nice-to-haves – they are already tools that allow us to be much more agile and much faster. Certainly, if we implement it right, if we trust that it works, and if we use it responsibly, AI can help us. The potential is limitless, we only need to look at our systems and processes to identify areas where AI can really benefit us. From our point of view, technology and AI will not replace us – they will certainly augment us and help us be more efficient.
Rareș Stănescu, PwC Romania Director: The biggest problems related to the adoption of AI that we see in the financial area are related to prompts and the analysis of results
It is clear that at the moment, when we talk about AI, we are no longer talking about a trend, but something that should be or is already in everyone's strategy. It helps us simplify processes, make them more efficient, have quick access to data and extract insights from that data, to facilitate quick business decisions. I am convinced that most of you already have access to those assistants that we see in Microsoft, in the Office suite or SAP. For example, we use it in the due diligence area, when we quickly extract certain clauses from the contracts and start comparing them, after which we identify some red flags. The effort saving is significant. Also, complex ERP systems and document management systems give us access to these agents or assistants who can make it easier for us to find documents.
We have implemented a solution called CFO Advisor, which helps you do a deep-dive analysis on company data. It connects to both the Internet and the company's databases, separates that information, and you can do financial analysis and go into the what-if zone – countless situations that you can explore. Everything is done with a suite of agents working together, not just one agent. This is the area we are moving into – how agencies will work together.
The biggest adoption issues we see in finance is that the technology is there, it works, but we're not yet addressing it the way we should. Prompts are a problem – the way we interact with technology on the one hand, and the way we analyze the output on the other. Most people say it doesn't work. In fact, technology works – it's how you relate to it that matters.
Robert Gîrdoc, PwC Romania Director: There is a very high risk that AI projects get stuck or are implemented without control due to governance issues
I think it's worth looking at not just what we can do with AI, but how we use it. The risks are much more varied – from bias in decisions, to hallucinations which are information that seems perfectly credible but which in reality are false, the lack of interpretability at the level of these systems, and perhaps the most important and subtle from my point of view is the tendency of people to sometimes delegate too much to these systems, without questioning the results or outputs they generate – which can cause very high exposures, both on a personal and professional level. All of these can affect a company's reputation and financial exposure very quickly. That's why I say that a responsible framework is not an obstacle – it's the foundation that makes innovation sustainable.
We see a lack of clarity in governance and a gap in understanding. Management wants to adopt AI, but doesn't know what questions to ask. Technical teams understand the patterns, but they don't know what the compliance risks or reputational risks are. And if we reach a context where business wants speed, and people from risk and compliance demand caution, there is a very high risk that projects will get stuck or be implemented without control. On the other hand, we see an exponential growth of the phenomenon of shadow AI – many internal employees using models based on artificial intelligence on their own, increasing the degree of exposure by entering sensitive, confidential data, customer data, financial reports, through these tools without control.
The AI Act requirements are very broad – from the level of data quality and relevance, how these models are overseen, the technical security measures we implement, how they are inventoried and managed. A lack of control over these aspects can result in very large fines – up to €35 million or 7% of global turnover, whichever is greater. And unlike the GDPR, where many companies waited to see how it is applied, here the signal is clear: early preparation is not optional, it is strategic.
Article supported by PwC Romania
