KPMG Hungary Reveals How AI Could Revolutionize Businesses

KPMG Hungary's Digital Audit Center also celebrates a milestone, having been established 10 years ago as the first business unit to introduce agile methodology.

“Artificial Intelligence (AI) is no longer knocking on our door – it has broken it down completely. By early adoption of modern technologies, we can gain a competitive edge,” emphasized Tamás Kórász, partner at KPMG Management Consulting.

According to Kórász, through digitalization and the development of Cloud, Big Data, and AI, Hungarian companies can not only advance but also open entirely new avenues for enhancing consumer and employee experiences, creating new value propositions, and transforming business models. While it remains uncertain how much AI will disrupt the business world and which professions and fields it will reshape in the next 10-20 years, it is already clear that AI can make a professional middle manager's work 40-50% more efficient within a few years. Although professional colleagues will still be needed, job roles and tasks will significantly change in all areas.

In the realm of AI, modern technological support is essential for business success, and KPMG forms strategic partnerships with the largest IT companies to achieve this. According to KPMG experts, proper functioning of AI requires dedication and experimentation.

AI Becoming Part of Daily Business Life

Ágnes Rakó, partner at KPMG Risk Consulting, highlighted that AI has already infiltrated traditional fields and is part of daily business life.

“AI allows for even more efficient use of existing human capacities, data, and resources. The use of AI is often an expectation from clients; it is a hallmark needed for products and services,” she noted.

Today, business processes and IT can only evolve together, with technology and business interests successfully collaborating using the same forward-looking tools. KPMG notes that it is dedicated to assisting its clients in digital transformation. This involves assessing the business, operational, and digital success factors of each company, creating business transformation and technology plans, implementing and integrating developments, and then keeping systems up to date.

To demonstrate tangible results and value, KPMG experiments within its own organization, allowing its own colleagues to see the value of AI in business use.

KPMG will soon publish the results of its ongoing, non-representative survey (Artificial Intelligence Maturity Survey). Preliminary data already indicates that most companies have tried AI solutions in some areas, and most have tried several tools, but they do not blindly trust the results generated by AI.

Peter Ignácz, KPMG's data analysis, modeling, and machine learning expert, also recommends this approach of data verification. He demonstrated through KPMG's deep learning system how machine learning and large language models can be used to extract relevant information from masses of documents and generate answers to posed questions, creating concise summaries of issues without human intervention.

“The corporate environment differs mainly from tasks solved with ChatGPT in that it does not use publicly available internet knowledge bases, but large language models process the company's own data, helping to solve their business problems,” emphasized Ignácz.

KPMG's Own AI Experiences

KPMG's own document processing system is already capable of formulating a large number of relevant questions for the desired processing of documents. The basis of the system is a well-posed question, as machine learning-trained systems can provide nearly 100% correct answers to well-posed questions, but like humans, large language models will only give wrong answers to poorly posed questions.

When running its own system on financial reports, KPMG found that it provided 80-85% usable analysis almost immediately compared to what two colleagues could have achieved with higher accuracy in about two days. Although the first run does not yield perfect results, it works excellently as a foundation that can be brought to 100% with human intervention in a fraction of the time, freeing up significant time for employees to focus on other tasks.

The application of machine learning and large language models is well suited for many corporate tasks, freeing up many human work hours for other tasks. However, for future-oriented technologies to be accurate, they need thinking and creative experts who precisely define the scope of input documents, tasks to be performed, and the expected content and accuracy of responses. In this area, the machines and humans do not compete but rely on each other to achieve leaps in productivity and responsiveness.