UiPath assisting One NZ on journey to become AI leaders

As the uptake of automation in workplaces continues to accelerate, a clear distinction must be made between consumer AI tools used for personal productivity and enterprise-grade AI deployments that influence business operations and customer outcomes.

While generative AI platforms such as ChatGPT and Perplexity have proven valuable for individual productivity, organisations require far greater oversight when AI is used to support business decisions, customer interactions or large-scale operational processes.

The concern centres on the growing risk of "shadow AI" – unsanctioned AI tools and applications being adopted across organisations without appropriate governance. 

This phenomenon mirrors the proliferation of unmanaged spreadsheets and macros that previously created operational risks for businesses, according to VP, Pricing & Product Strategy at UiPath, Feiran Hao.

"For personal productivity, it's incredible," Hao said at the UiPath Fusion Summit in Sydney.

"I'm a big user of it, I think it's great in certain areas.

"But I certainly don't think of using ChatGPT and Claude, and all these kind of personal productivity tools for enterprise orchestration work, because enterprise work is where you need some sort of governance, some sort of control."

Unlike traditional AI systems, which typically deliver consistent outputs from the same inputs, large language models can generate different responses to identical prompts. This introduces challenges around auditability, accountability and trust, particularly when AI systems are deployed across critical business functions.

This is why business orchestration is rapidly coming to the fore as a key enterprise solution that helps organisations better manage how AI and legacy systems work together, and supports improved governance and visibility.

For New Zealand telco OneNZ, which aims to become a leader in AI-enabled telecommunications, governance has become a central component of its AI strategy. 

It has established strict controls governing how AI tools can access corporate information, with different classifications of data determining whether employees can use generative AI platforms.

Highly sensitive customer information and commercially confidential data remain subject to stringent restrictions.

Cy Wright, General Manager at One New Zealand, explained that the company has deliberately avoided allowing employees to independently build AI agents that connect to critical enterprise systems. 

"We have data engineers and AI engineers working with process people to build those agents in a way that we know that a small number of people build and many eyes validate. 

"Because if you've tried to use agents across finance data, you can get some really bad answers fast, and the problem is, the people who have asked the question, a lot of the time don't actually know that the answer is not right."

The approach reflects concerns that poorly governed AI agents could generate inaccurate outputs or unintended business consequences, particularly when operating across complex financial, billing or customer management systems.

One example highlighted by One NZ involved the use of AI-powered controls to verify that negotiated customer pricing agreements are accurately reflected in billing systems. Such applications require access to highly sensitive data and therefore demand extensive oversight and validation.

To support these initiatives, it has adopted a multi-platform strategy that incorporates technologies from cloud providers and automation vendors. Business orchestration is a key component of this strategy. 

Identity management, access controls and single sign-on capabilities are also critical components in ensuring employees can only access information appropriate to their role.

Emphasis on security reflects the realities of operating in New Zealand's highly competitive telecommunications market.

A significant customer data breach could have severe consequences for market share, given the relative ease with which consumers can switch providers. In a market of approximately 5 million people, maintaining customer trust remains a strategic priority.

Beyond governance, One NZ has harnessed AI to improve productivity and address workforce challenges.

The firm has launched employee training initiatives designed to increase AI literacy across the organisation and help staff understand both the opportunities and limitations of emerging technologies.

The company previously framed AI as technology that would support employees rather than replace them. 

The firm is also exploring how AI can help employees transition into new areas of demand, particularly in data, engineering and AI-related disciplines.

Instead of focusing on workforce reduction, the company is exploring how AI can help employees transition into new areas of demand, particularly in data, engineering and AI-related disciplines.

This is especially relevant in the NZ market, where businesses often struggle to recruit specialised technology talent.

AI staff training could allow the organisation to develop capabilities internally, potentially reduce reliance on offshore service providers and systems integrators.

According to Wright, in some cases, AI may even enable work currently performed overseas to be brought back in-house, creating new opportunities for local employees while reducing operational costs.

"We're trying to empower our employees as much as possible to get comfortable with the AI, but it depends on the individual. Everybody is different.

"For us, we try to get our employees on board (with upskilling). For example, full stack coders, it can be a struggle to find these people in New Zealand.

"What we want to be able to do is identify those people with some of those core skills and we have a whole lot of training programs that people can go online and do internally."

Despite optimism about AI's potential, there remains uncertainty around how workforce models will evolve over the coming decade.

Many telecommunications roles involve highly structured, repeatable processes that are well suited to automation. As AI becomes more capable, the nature of these positions is expected to change substantially, although the ultimate impact remains difficult to predict.

Hao is optimistic about automation and business orchestration becoming tools humans can leverage to improve efficiency, as opposed to making people obsolete, when addressing concerns around so-called dark factories.

Rather than viewing automation as a mechanism for eliminating jobs, UiPath believes the greatest opportunities lie in removing repetitive and low-value tasks that employees typically dislike performing.

"The areas where we look at things going towards dark factories isn't so much about trying to take away jobs. It's really trying to automate areas that people traditionally didn't like to do," Hao said.

"I don't think anyone actually liked testing. People liked thinking about testing and strategising around testing and making sure that use cases and edge cases were covered. But the actual work of click, click, click, did that work? Click, click, click, did that work? That is kind of mind-numbing work.

"Let me go automate them instead and run that; did that work or not?"

UiPath's Maestro orchestration platform, which now combines traditional business process management techniques with more adaptive case-management capabilities, is an example of how business orchestration improves automation.

Organisations increasingly require workflows that can respond dynamically to changing information, instead of following strictly linear paths. This flexibility is becoming increasingly important as AI is integrated into customer service, insurance, lending and other complex operational environments.