Earlier this year, PwC New Zealand welcomed a new team member to their workplace: a GenAI tool known as ChatPwC. Developed in partnership with Microsoft, it uses OpenAI GPT and Microsoft Azure OpenAI Service to respond to text-based questions within a secure PwC environment.

Across the global PwC network, PwC New Zealand is the first member firm in Asia Pacific and the fifth worldwide, to integrate this AI technology into its wider workplace tools. The organisation has also forged alliances with other major AI technology vendors, including Amazon Web Services (AWS), Anthropic, Google, and Meta.

Kayur Patel CA, senior manager, emerging technology and GenAI at PwC New Zealand, explains that by positioning PwC as ‘client zero,’ the firm is transforming its own operations at scale and is better positioned to help clients do the same.

“We knew that we had to do this ourselves first, so that we could speak from experience when we took it to the market.”

How is client data protected?

As an enterprise-level ChatGPT AI solution, ChatPwC ensures that PwC client and company data is protected behind a firewall. Client data is never reused or stored, and all information remains confidential.

Patel, who hosted a panel at AFAANZ 2024 on the future of talent, technology, and GenAI, explains the uses range from assisting with document drafting to industry analysis and checking work.

“Essentially, anytime you’ve got a large body of text that needs to be created, distilled or reviewed, those are the use cases that we apply at the moment,” he says.

What were the benefits of introducing GenAI?

PwC’s latest global CEO survey shows 68% of CEOs in the US (64% globally) expect GenAI to increase employee efficiency this year and 44% expect it to increase profitability.

PwC in the US also reported productivity gains of its own since investing in AI. For example, its IT department has experienced 20–50% productivity gains in software development processes, and finance has seen a 20–40% productivity gain in accounting and tax.

Patel explains that the benefits go beyond productivity.

“When we’re advising our clients about this kind of transformation, we can speak from our own experience,” he says, adding that the quality of work has also lifted.

“If we use this technology well and have it access the repository of PwC intellectual property, we can start to really improve how we standardise work and improve the quality.”

Were there any barriers?

The implementation of AI tools requires a careful onboarding process.

“Unless you can help people identify the use cases that will give them those gains in productivity and quality, it’s never going to be that successful,” says Patel. “You have to go through the whole change process.”

Leadership buy-in is also essential to success.

“You can’t get your leadership to just sign off on it and get the experts to run it,” says Patel. “We had our full executive leadership team on board. We trained them on how the technology actually works and to understand the impact it would have on the business, the industry, and the use cases.

“Once we did that, we created an AI strategy and risk policies, and then ran a change management process and a pilot program.”

Checks and balances

Patel explains that quality control is a key consideration for the use of AI tools.

“We’ve got a list of the things that we can and cannot use it for, and the different pieces of data that we can and cannot put in,” he says. “We’ve gone through a stringent review process with our privacy and legal experts to cover this.”

Patel adds that “nothing leaves the door without a partner reviewing it.”

“This has always been the case, whether or not we’ve used AI to help create a piece of work. If it doesn’t meet the threshold for a partner’s review, it’s not going to go out.”

Patel notes that the change management process is ongoing.

“We’ve got what we call ‘supercharging sessions’ to help our teams understand how to build custom assistants, instead of just using it in a prompting fashion,” he says.

“As we identify more things that the technology can help with, we are raising the bar, and that happens as we get access to more and more models. But, with any model that you use, it’s important to identify the line between where its innovation ends and where yours starts.

“If you’re not careful about that line, you’ll find that, in a few months’ time, the things that you are doing to innovate will just come out of a box,” adds Patel. “We've been really careful about understanding what we are going to take from a model and what we are going to build around it to help our people.”

---

AFAANZ: improving accounting and finance education and research

The Accounting and Finance Association of Australia and New Zealand (AFAANZ) aims to promote excellence in accounting, finance and other related subjects through the development of teaching and research in Australia and New Zealand. Its annual conference seeks to advance this mission and to showcase the skills and contributions of its diverse membership, which includes academics from the higher education sector, Australian and New Zealand universities, accounting and finance industry professionals, and corporations.

AFAANZ 2025 will take place 6–8 July 2025 at Sofitel Brisbane Central, Queensland.

---

What is the meaning of GenAI?

Short for generative artificial intelligence, GenAI comprises artificial intelligence (AI) systems that are capable of generating new content, ideas, or data that mimic human-like creativity.

What is GenAI vs ChatGPT?

Think of GenAI as an overarching term that ChatGPT sits under. While GenAI is a broad field of artificial intelligence, ChatGPT is a specific implementation of it.

What are GenAI examples?

Thanks to its advanced capabilities, GenAI can significantly improve supply chain processes by analysing unstructured data and extracting relevant information. This means it can do things like help businesses evaluate supplier performance, negotiate better terms, and identify potential risks or bottlenecks in the supply chain.

What is generative AI vs AI?

Unlike traditional AI systems, which are programmed with specific rules and algorithms, generative AI systems are trained on large datasets.

---