Artificial intelligence can be helpful, but it remains a tool

A Tax Control Framework in a snap with GenAI

  • Blog
  • 05 Dec 2024
Marvin de Ridder

Marvin de Ridder

Director, PwC Netherlands

Daan Meerdink

Daan Meerdink

Senior Manager, PwC Netherlands

ShapeGenerative Artificial Intelligence (GenAI) is on the rise, offering organisations new opportunities. Marvin de Ridder and Daan Meerdink, experts in tax control at PwC, question whether GenAI could be the key to successfully establishing a Tax Control Framework (TCF).  

GenAI is currently a hot topic, capable of creating new content such as text, images, videos, audio, or software code. As consultants, this raises the question: to what extent can GenAI facilitate or even improve the setup of a Tax Control Framework? 

To ensure an organisation adequately manages its tax obligations, it is essential to establish a Tax Control Framework (TCF). Setting up such a framework involves several components, including formulating a fiscal strategy, mapping and documenting roles, responsibilities, and processes, identifying risks, formulating control measures, and creating a monitoring and testing plan.

  1. Tax strategy
    The tax strategy is a document where an organisation lays the foundation for the TCF by setting its tax objectives and how it plans to achieve them. This includes risk appetite, the role of the tax function, the desired relationship with the tax authorities, and the involvement of the board. By providing data about the organisation, such as its general mission, vision, and strategy, GenAI can quickly create an initial draft of a tax strategy.
  2. Roles and responsibilities
    For a mature TCF, it is crucial to have clarity on the distribution of tax responsibilities and the involvement of various parties. A so-called RA(S)CI matrix can help by displaying tasks and stakeholders in an overview. GenAI can generate a RA(S)CI matrix based on existing documents and texts that describe individuals and their roles.
  3. Tax processes
    Documenting tax processes can be made visual with flowcharts or extensively explained in process descriptions. Most GenAI tools can analyze existing documentation, such as manuals, and distill current practices from them. If no document is available, GenAI can also create a questionnaire for employees to fill out and then generate a description based on the obtained answers.
  4. Risk management
    Effective tax risk management is an essential part of a TCF. By managing its risks, an organisation can ensure it achieves its broader business strategy and objectives and gains the trust of the tax authorities and other stakeholders. Risks and control measures are typically identified by inspecting relevant documentation and interviewing various stakeholders. This information is then compiled into a risk and control matrix (RCM). While GenAI cannot yet fully automate this process, it can assist in preparing interview questions, analyzing transcripts, and quickly gathering data from other sources, such as process descriptions and reports.
  5. Monitoring and testing
    Monitoring and testing involves regularly checking whether the tax processes and control measures are effective enough to mitigate identified risks and whether there are any unknown risks. Ideally, this is also reported to stakeholders, such as the board and possibly the tax authorities. To do this in a structured manner, a monitoring plan is needed. This plan should describe the purpose, scope, responsibilities, reporting, and follow-up of monitoring and testing activities. Activities are displayed in a calendar with tax resources, linked risks, executors, and testing periods.  

AI can help set up such a plan by gathering data from the aforementioned TCF documents and making recommendations for suitable monitoring activities.  GenAI can also be used to articulate the results of various monitoring activities in reports on the progress of these activities.

The importance of human involvement with AI

There are many ways in which GenAI can assist in setting up a Tax Control Framework.  However, there is a caveat. How practical is the generated content, and to what extent can you trust the outcomes? The way you formulate a 'prompt,' the written command given to the AI tool, significantly influences the answer. GenAI models are trained on human language. 

A broadly formulated question like 'create a tax strategy' will likely yield an answer, but often not the desired one. Additionally, the answer may be based on training data that might not be up to date with recent regulations or the latest version of certain standards. 

Specifically formulating the prompt and including relevant criteria as a framework, such as guidelines from the Dutch Tax and Customs Administration's Supervisions of Large Businesses in the Netherlands or the VNO-NCW Tax Governance Code, will lead to a much more accurate result.  Some prior knowledge is therefore necessary to verify the accuracy of an answer and to properly guide GenAI. 

There can also be a so-called 'bias'. This mainly occurs if the prompt steers in a certain direction.

A Tax Control Framework in a snap with GenAI

Responsible use of GenAI

Governments have started implementing regulations for the responsible use of AI. Since August 1, 2024, the EU AI Act has been in effect.  This European regulation applies to anyone who creates, uses, imports, or distributes AI systems in the EU, regardless of where they are based. Its goal is to protect EU citizens with uniform standards while promoting value creation in the field of AI. The UK tax authority, HMRC, also emphasizes the importance of responsible use of artificial intelligence. The HMRC has recently published guidelines for implementing Tax Control Frameworks. These guidelines highlight several points for the use of artificial intelligence in activities such as processing transactions and assigning the correct tax codes. For example:

  • Tax decisions that rely on algorithms should be explainable in terms of decision factors and audit trail and monitored for accuracy.

It is challenging to trace the source of information used by GenAI to generate content. An answer may also contain content protected by copyright, leading to unintentional plagiarism. Conversely, there is a risk that users share sensitive information, which then becomes part of the training data and accessible to third parties, especially with free or public AI tools. 

Developing awareness and skills to use GenAI safely

Proper training to effectively and safely use GenAI can help avoid the mentioned drawbacks. Think of courses on writing good prompts and being aware of AI pitfalls. In short, while GenAI can be a powerful tool for improving the technical and procedural aspects of a TCF, the human factor remains essential for its success.

It is therefore important to consider the following aspects:

  • Be as clear as possible when writing a prompt and include the framework explicitly.
  • Be aware of different types of biases that may occur.
  • Be aware not only of the potential that GenAI offers but also of the data risks and regulations regarding its responsible use.
  • Continue investing in maintaining the knowledge level within the organisation so that employees can develop the skills to leverage GenAI capabilities and ensure the quality of the generated content.

Want to know more about setting up a Tax Control Framework correctly?

About the authors

Marvin de Ridder
Marvin de Ridder

Director, PwC Netherlands

works within the tax practice of PwC Netherlands and is based in Rotterdam. He has over twenty years of work experience in various fields, such as finance, controlling and tax risk management.
Daan Meerdink
Daan Meerdink

Senior Manager, PwC Netherlands

works within the tax practice of PwC Netherlands and is based in Amsterdam. Within the practice, he focuses on tax strategy & operations.
Follow us