The rapid development and adoption of artificial intelligence (AI) is transforming the way businesses choose to operate. Many organisations are turning to platforms like ChatGPT to streamline day-to-day processes, and The Times reports that nearly a third of UK employees are using AI at work.
Payroll is often one of the most time-consuming aspects of company operations, so it is unsurprising that people are increasingly using AI to assist with payslips, pension schemes, workforce cost tracking and other administrative tasks.
When used effectively, AI can offer advantages for businesses aiming to improve their payroll, helping to input large amounts of data quickly and at scale. As a result, 42% of companies are opting to use AI in payroll, according to HR News.
However, this requires caution. Without regulated checks by professionals, mistakes ranging from AI hallucinations to automation errors can go undetected. This has a real impact on people and trust, not simply operations. Now more than ever, businesses must be careful to avoid issues like data breaches, reputational damage or a lack of compliance to protect their employees and values.
Payroll is an essential part of any organisation. It ensures that employees receive the pay and benefits they are entitled to fairly and on time, while enabling companies to operate efficiently and in accordance with legal regulations.
For account keeping and finances, attention to detail is paramount, and AI can help to make record-keeping smoother while removing the potential for human error. Systems like ChatGPT or Google Gemini can collect and organise high volumes of data, identifying trends and making predictions to support long-term business planning. They will also have some knowledge about HMRC tax standards, National Minimum Wage regulations and other statutory payroll requirements, but not to the extent that a seasoned professional or a regulated AI payroll system will.
As large language models (LLMs), these AI systems are not designed for processing numerical data and shouldn’t be relied on for financial compliance. ChatGPT can excel at language processing by generating text based on learned patterns of words, but they struggle with complex, multi-step or multi-digit mathematics because they focus on pattern recognition and probability, rather than any real understanding of maths.
Put simply, Gemini, ChatGPT and CoPilot may help in some circumstances, but they aren’t the right tool for proper payroll processing.
There are other types of AI which are better suited for data processing and mathematical problem solving, such as Machine Learning (ML) and Intelligent Document Processing (IDP) models, but even these carry risks when it comes to data security, legal compliance, errors and hallucinations.
Payroll operates in a highly regulated environment that requires consistent monitoring, so businesses considering greater reliance on AI must understand its risks, with one of these being data breaches.
Businesses are responsible for storing a multitude of sensitive information, including employee salaries, National Insurance numbers and bank account details, so introducing AI into payroll processes can complicate this by risking GDPR and privacy breaches.
This risk comes from the AI provider’s cybersecurity systems that may be vulnerable to data breaches. And also, many AI models are designed to learn from the data they are given by default, meaning sensitive payroll information can be stored, processed or reused in ways that create additional compliance and privacy concerns.
Without appropriate safeguards, disclosing company and employee data may increase exposure to cyber threats and security risks. To handle data appropriately, it is important for businesses to use secure, encrypted platforms with controlled access. Professional accounting services use AI systems that are GDPR compliant to protect data integrity, giving employees and employers peace of mind that their data is safe.
Beyond data protection, employers must meet wider legal standards such as PAYE regulations around tax, National Insurance, student loans and pension enrolment. The margin for error is narrow, while the consequences for people and finances can be significant.
Although AI can assist with providing legal information, it cannot replace professional judgement or guarantee that ever-changing legislative changes are applied correctly. Outsourcing can remove this complexity and help companies to navigate policies without needing an in-house team, freeing up valuable time in other areas.
From regular operational support to corporate consulting, professional payroll services prioritise accurate and strategic advice, protecting the reputation of businesses with up-to-date insights and experienced advisers. This kind of safeguarding cannot be secured by AI alone.
Automation can be implemented in the payroll function to carry out repetitive tasks such as calculating pay, applying deductions and generating reports. This can make the payroll process simpler, but it also means that any errors can repeat automatically across large data sets - and employers will be held to account.
Even small errors, like a missing digit, can cause major issues, so it is vital for automation to be under effective control. This highlights the importance of combining AI with professional review to protect employees and employers.
Technology should enhance efficiency, but not at the cost of accuracy and protecting people within a workplace. Taking a balanced approach through professional monitoring helps prevent technical errors from becoming costly and concerning problems.
AI hallucinations have come under increasing scrutiny, exposing some of the shortcomings of all kinds of generative AI. They occur when AI systems produce responses that seem convincing but are inaccurate or outdated, as they are designed to predict likely responses.
Within payroll, this is particularly concerning. National standards are likely to change on a regular basis, so relying on unverified information is a key risk factor for businesses. Widely accessible AI tools, like ChatGPT, are not subject to the same standards as regulated software with input from payroll professionals.
They can also be trained on unreliable information, with systems struggling to distinguish whether the information from different sources is accurate and trustworthy, and it has been demonstrated that AI systems can be manipulated by false information published online.
For business leaders, the message is clear - AI can help streamline payroll processes, but the type of AI needs to be suitable for the task, and it shouldn’t be used as a standalone strategy.
Calculations have to be accurate and reliable, but payroll is also centred around people, trust and responsibility. Although AI can sharpen the process, it is not a substitute for professional insights.
Mistakes in payroll can be extremely damaging for any company, so payroll has to be handled with care and attention to detail. When using AI, some of that is taken away, so it is always better to take a more cautious approach and rely on the services of a dedicated payroll professional.
Duncan & Toplis provides payroll services for businesses across a range of sectors, including payslip processing, pension scheme management and workforce cost tracking. To find out more, contact us.
